obra

I wrote a code-graph plugin for Claude Code and tested it against three popular ones on a real task. None of them helped. Every Claude Code session starts the same way. The agent does not know my codebase, so it greps, reads files and guesses what a change will break. I thought it would be great to give it a map. The agent asks one question, "what does this change affect?", and gets a short list of files and tests. The map is built from imports and from the contracts between packages: - HTTP routes - GraphQL operations - tables - queues - env variables So I built that tool and called it gsttt. A week later I had tables and charts in which it beat three popular tools. Then I tested all four on real work, and none of them made a measurable difference. This post is the whole story, from the idea to the moral. It includes the test setup, so you can check my work or redo it on your own repository. The short version - The idea: a map of the code should help an agent find what a change affects. - The tool: I built gsttt, and my own benchmark said it beat Graphify, GitNexus and codebase-memory-mcp. - The first reality check: on nine real commits, the installed tool was never called: 0 of 36 runs. - The last try: on one large real task, all four tools cost the same as no tool, inside the noise. - The moral: the agent did not need the graph. The bill followed which model does the code reviews and how many review rounds run. Contents - [The idea and the build](#the-idea-and-the-build) - [The first reality check](#the-first-reality-check) - [One last try with a real task](#one-last-try-with-a-real-task) - [The result](#the-result) - [Did the agents use the tools](#did-the-agents-use-the-tools) - [What moves the bill](#what-moves-the-bill) - [The moral](#the-moral) - [Limits](#limits) - [If you want to reproduce it](#if-you-want-to-reproduce-it) The idea and the build The code-graph tools I found did not treat contracts between packages as first-class edges. A GraphQL field in the server and the client query that uses it are an example. I wanted that link in the map, so I wrote my own on October 28. It is written in Rust, mostly with Claude Code. To see if it worked, I built a small multi-language fixture repository. I added fourteen questions with truth lists, such as "what depends on this file". Mean recall: | Tool | Mean recall | |---|---| | gsttt | 87% | | GitNexus | 73% | | codebase-memory-mcp | 72% | | Graphify | 46% | On Amplication, an open-source TypeScript monorepo, the TypeScript compiler also listed 97.3% of the callers gsttt found for 200 random methods. I was pleased. Then I looked at what the benchmark measured. I had written the fixture and the questions, knowing what gsttt does. A benchmark you write for your own tool measures whether it does what you built it to do. The first reality check So I ran the tool on real work: nine real merged changes, five from Amplication and four from a private monorepo. Each agent got a ticket written from the commit message, and I compared its diff with the real one. The model was Sonnet 5.5 at high effort, 99 runs in all. - Installed only: the agent never called gsttt. 0 of 36 runs, and 0 in 107 real sessions of mine. - With one sentence in the ticket ("Use gsttt to find what the change affects."), it called gsttt in all 27 runs. - The result: 4.4 more turns, about 4% more money, and no better recall of the real files: 0.64 against 0.66. I tried to fix that: - a skill text that says when to call the tool - answers in functions instead of files - cheaper answers - hints placed in the planning step and at stop time Nothing helped beyond the noise. An offline check on 420 commits compared gsttt with git co-change history. At the same list size, git history found 13 points more of the real files. For these agents, the graph carried no information they could not get from search. But those were small commits. A large change across many packages might be where a map pays off. One last try with a real task I also wanted the other tools in the picture. They have between 48,000 and 142,000 GitHub stars, and gsttt has 0. So I picked one large real task and ran all four tools on it. The task is a merged change in a private TypeScript monorepo for a SaaS product: an annual billing interval with a discount. It touches: - price calculation, the payment integration and the checkout service - a GraphQL contract on both sides, regenerated into three packages - a plan picker, a pricing section and translations in five locales - tests and an OpenSpec change proposal The real diff is 36 files, +1,183 and -96 lines. I chose it from 377 recorded sessions in that repository. It was the cheapest of four that edited at least eight files across three packages without any tool. The original session used Claude Opus 5.5 at xhigh effort. It took 57 turns and 15 sub-agents, cost about $8.10 and also made commits and a pull request. The codebase has six packages and 2,809 tracked files, about 282k lines of TypeScript. It carries its own agent setup, which stayed in every run: a CLAUDE.md, two sub-agent definitions and nine skills. The setup: - Each run starts in a fresh git repository with one commit, the parent state. The agent has no network and cannot read outside its folder. - The ticket is 335 words. Every run used Claude Opus 5.5 at xhigh effort, Claude Code 2.1.318, with a $15 cap. - Plugin only: the tool is installed as its README says, and nothing else. - Enforced (bare rounds only): the ticket also demands a tool call before the first edit and before the end. The gate never fired, because the agent called the tool first anyway. Two environments. The first one was a mistake. - Bare (rounds 1 and 2) loaded only the repository's own files. It is clean, but I never work that way, so its dollar figures do not describe my sessions. - Harness (round 3) also loads my own setup: my agorch plugin and superpowers, the two plugins of the original session, plus my instructions and rules. - In both, nothing else is loaded and nothing reaches outside the machine. Every run's start is audited against its arm. The tools are these four, at the versions I ran: | Tool | Version | Licence | Stars | |---|---|---|---| | gsttt | 0.1.196 | Apache-2.0 | 0 | | CodeGraph | 1.6.2 | MIT | 77k | | Graphify | 0.9.102 | Apache-2.0 | 142k | | GitNexus | 1.6.12 | PolyForm Noncommercial | 48k | I am the author of gsttt. No index build calls a model. The checks are mechanical: - typecheck of three packages - the server and client test suites - the GraphQL schema compared with the committed one - nine held-out test files from the real change - a 16-item checklist from the real diff, judged blind The result Cost is total_cost_usd at list prices, sub-agents included. | Setup | Runs | Cost per run, USD | Mean | |---|---|---|---| | no harness, baselines | 3 | 10.72, 11.74, 11.84 | 11.43 | | no harness, plugin only (four tools) | 4 | 11.75, 12.37, 11.25, 11.28 | 11.66 | | harness, baselines | 5 | 6.84, 7.39, 7.83, 8.73, 9.28 | 8.01 | | harness, gsttt | 2 | 9.44, 9.31 | 9.38 | | harness, Graphify | 3 | 8.19, 10.29, 8.54 | 9.01 | | harness, CodeGraph | 2 | 7.75, 9.55 | 8.65 | | harness, GitNexus | 2 | 7.40, 9.72 | 8.56 | !Dot chart of the cost per run in USD for no tool, gsttt, Graphify, CodeGraph and GitNexus, without and with my harness. In both panels the tool runs fall inside or next to the spread of the baselines. The full-size chart Quality did not move Every run delivered the whole change: - File recall against the real diff is 1.00 (0.97 once). - The checklist is 16 of 16. - The typechecks are clean and the GraphQL schema equals the real one. The nine held-out tests failed in every run, because they pin names the original author chose. The unaided agent already solved the task, so a tool had nothing to add except savings. The savings are inside the noise The three bare baselines differ by 1.12 dollars, so a difference of about a dollar cannot be read. The harness baseline mean is 8.01. | Tool | Runs | Difference from baseline mean | |---|---|---| | gsttt | 2 | +1.36 | | Graphify | 3 | +1.00 | | CodeGraph | 2 | +0.64 | | GitNexus | 2 | +0.55 | None is distinguishable from the baseline. The same tool run twice differs by 0.13 (gsttt) to 2.32 dollars (GitNexus). In the two big gaps, one run had one review round and the other had two. gsttt has the highest mean, and both its runs had two review rounds. I wrote this tool, so weigh that. Did the agents use the tools Maybe the agents just ignored them. They did not. In every plugin-only run the agent used its tool unasked, because of what each installer writes into the project. - Graphify adds a CLAUDE.md rule to query it before the first grep, two hooks and a 100 KB skill. The agent's first command was a graphify query. - CodeGraph adds an MCP tool and a hook that injects 9,223 characters of context at the first prompt. The agent called codegraph_explore twice, each answer about 23k characters. - GitNexus adds about 45 lines of MUST and NEVER rules, 12 skills and two hooks. The agent called impact 12 times, and the hook added a note to 10 of 90 Bash calls. - gsttt writes nothing into the repository. The agent called impact three times. The CodeGraph run was the most expensive of the twelve bare runs, $12.37. I cannot say what drove it. The gsttt calls contradict my earlier 0 of 36 for installed-only gsttt. The model, the task and the plugin version differ, and one run per tool is not a rate. So the tools were used, and the agent still did not need them. What moves the bill If the tools do not move the cost, something else does. The harness makes the same task $3.42 cheaper, 30%. My instructions send review and builder agents to Sonnet. - In all 43 review launches of the 14 harness runs, the agent passed model: "sonnet". Without the harness, 34 of 40 ran on Opus. - No experiment changed only the reviewer model, so this is a reading of the numbers. The spread inside the harness is wider, 6.84 to 9.28 among the baselines. It follows the review rounds the agent runs: | Review rounds | Baselines, mean USD | Tool runs, mean USD | |---|---|---| | one | 7.35 | 7.90 | | two | 9.00 | 9.42 | I do not know why the agent runs one round in some runs and two in others. It moves the cost by about two dollars, more than any tool did. The moral On paper, this looks great. A graph of how your code connects is a fun thing to build, and it is fun to look at. Many people install tool after tool and believe that the more tools they add, the better their harness and their model get. I started from the same belief. On this task, Claude did not care about these cool graphs of the codebase. It found every file without them. What moved the bill was boring: which model does the reviews and how many review rounds run. So before you add the next tool to your harness, measure it on one real task of yours. Run the baseline at least three times first. If the number does not move, the tool is decoration. Limits - One task, one repository, one model at one effort. No statistical test. - Recall and the checklist saturated at the maximum, and one AI agent judged the checklist. The typecheck, the tests and the schema equality carry the correctness evidence. - Runs are few: the bare environment has 3 baselines and one run per tool arm, the harness 5 baselines, 2 runs per tool and 3 for Graphify. - In round 1, ten claude.ai connectors were registered by mistake. No connector tool was called. Hooks of Graphify and GitNexus may have run twice in their round-1 enforced runs. - The harness comes from one original session. I trust the direction, cheaper reviews and a wider spread, more than the exact figures. - Wall-clock time is not comparable, because two runs shared the machine. - I wrote one of the four tools and designed the test. The checks are mechanical, but weigh that. - A task where the unaided agent misses files would be a fairer test of a code graph. This one did not have that problem. If you want to reproduce it 1. Pick a merged change in your repository that spans a few packages. 2. Rebuild the parent commit as a single commit and write the ticket from the issue. 3. Run each setup with the same model and effort, in a folder the agent cannot leave, with your own setup loaded. 4. Install each tool as its README says, writing into the project and not into your global config. 5. Set CLAUDE_CODE_PRINT_BG_WAIT_CEILING_MS=0 for headless runs with sub-agents. By default the process stops background tasks 10 minutes after the last turn. 6. Run the baseline at least three times before trusting a difference. The test runs, the blind judging and the first draft of this write-up were done by Claude Code agents from the run logs.

Superpowers: The Open-Source Method That Stops Your AI Agent Rushing to Code --- title: "Superpowers: The Open-Source Method That Stops Your AI Agent Rushing to Code" published: true tags: opensource, aicoding, claudecode, productivity canonical_url: devpik.com --- Superpowers stops your AI coding agent from jumping straight into code. It asks what you actually want, plans step by step, writes tests first, and sends sub-agents to do each task, with a review after every one. It works with Claude Code, Cursor, Codex, Copilot, Gemini and more, and it has 309,623 stars, making it one of the most starred developer projects on GitHub. What Is Superpowers? Superpowers is a complete software development methodology for your coding agents, built on top of a set of composable skills and some initial instructions that make sure your agent uses them. It is built by Jesse Vincent and the team at Prime Radiant, licensed under MIT, and the repository had 309,623 stars and 27,550 forks when we checked on October 31, 2026. The core idea is simple, and it fixes the most common failure of AI coding. Left alone, a coding agent jumps straight into writing code, guesses at what you want, and produces something that sort of works and is hard to change. Superpowers changes the default. From the moment you fire up your agent, it steps back and asks what you are really trying to do, instead of trying to write code first. Because the skills trigger automatically, you do not need to do anything special once it is installed. Your coding agent just has Superpowers. The skills fire at the right moments: brainstorming before building, planning after design approval, test-driven development during implementation, and code review between tasks. They are mandatory workflows, not suggestions your agent can skip when it feels confident. Brainstorm First, Then Plan The workflow starts with brainstorming. The agent draws out what you actually want, and why, through questions that get you describing it, then plays its understanding back for correction. Small changes get agreed in chat. Bigger projects get a written design document that a builder could plan from, shown to you in chunks short enough to actually read and digest. After you sign off on the design, the agent puts together an implementation plan that is clear enough for an enthusiastic junior engineer with poor taste, no judgment, no project context, and an aversion to testing to follow. That is the project's own bar, and it is deliberately low: if the plan only works when read by an expert, it is not a good plan. Work is broken into bite-sized tasks of 2 to 5 minutes each, and every task has exact file paths, complete code, and verification steps. Then, once you say go, the agent launches a subagent-driven process. Fresh sub-agents work through each engineering task, with inspection and review of their work as they go. It is not uncommon for an agent to work autonomously for a couple of hours at a time without deviating from the plan you put together, because the plan, not the agent's enthusiasm, is what is driving. Test-Driven by Default Superpowers emphasizes true red-green test-driven development, YAGNI, which means you are not gonna need it, and DRY, do not repeat yourself. During implementation, the test-driven-development skill enforces the red-green-refactor cycle: write a failing test, watch it fail, write the minimal code to pass, watch it pass, then refactor. Code written before tests gets deleted. That sounds strict, and it is meant to be. AI agents are very good at writing plausible code that has never been run, and tests are the cheapest way to catch that. The verification-before-completion skill adds a second guard: the agent must ensure something is actually fixed before declaring it done, instead of claiming success from reading its own output. Between tasks, the requesting-code-review skill reviews work against the plan and reports issues by severity. Critical issues block progress. When tasks complete, the finishing-a-development-branch skill verifies tests, presents options to merge, open a pull request, keep or discard the branch, and cleans up the worktree. Nothing merges on vibes. The Skills Inside Superpowers ships a library of skills grouped by job. Testing has test-driven development with a testing anti-patterns reference. Debugging has systematic debugging, a four-phase root cause process, verification before completion, and diagnosing-superpowers, which works out what went wrong in a session with line-level evidence and can package a scrubbed bundle for a bug report. Collaboration skills cover the whole lifecycle: brainstorming, writing plans, executing plans inline, dispatching parallel agents, requesting and receiving code review, using git worktrees for isolated branches, finishing a development branch, and subagent-driven development with two-stage review, first for spec compliance, then for code quality. Meta skills cover writing new skills and the using-superpowers bootstrap itself. The philosophy underneath is four lines: test-driven development, write tests first, always. Systematic over ad-hoc, process over guessing. Complexity reduction, simplicity as the primary goal. Evidence over claims, verify before declaring success. If those sound like rules a good senior engineer would enforce on a team, that is the point. Superpowers is a senior engineer's habits, packaged as skills an agent cannot quietly drop. Works With Your Agent Superpowers is not tied to one tool. It installs as a plugin or extension for Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot CLI, and many more, including Antigravity, Devin, Factory Droid, Grok Build, Kimi Code, OpenCode, Pi, Qwen Code, Hermes Agent, and Muse itself. Installation differs by harness, and if you use more than one, you install it separately for each one. For Claude Code, install it from Anthropic's official plugin marketplace with /plugin install superpowers@claude-plugins-official, or register the Superpowers marketplace and install from there. For Cursor, install from the marketplace in Agent chat. For Codex, find it in the official plugin marketplace. The README lists every harness with its exact command, so pick yours and follow that one. One detail worth knowing: skills trigger automatically, but the bootstrap that makes them trigger is injected at session start. If a very long session compacts and skills stop firing, start a fresh session. On Hermes Agent, which has no post-compaction hook, the same applies. And if a session misbehaves, ask your agent to figure out what went wrong with Superpowers in that session, and the diagnosing skill will read the transcript and report what happened with evidence. How to Try It Pick your coding agent and install the plugin. In Claude Code, run /plugin install superpowers@claude-plugins-official. Then start a fresh session and describe something you want to build. A working install auto-triggers brainstorming before any code is written, so the first thing you should see is questions about what you want, not code. A good first test is a small project with a real requirement, like a todo app with a specific behavior, or a small feature on a codebase you already have. Watch for the sequence: questions first, a design played back for correction, a plan broken into small tasks with exact files and verification steps, tests before code, and a review after each task. If your agent skips straight to code, the install is not active, and a fresh session usually fixes it. Updating is often automatic, but it depends on your harness. The project also offers commercial services for enterprises that want support, additional tooling, or managed spending, through Prime Radiant. For everyone else, the open-source version is the whole methodology, free under MIT, and it is what the 309,623 stars are for. What to Watch Out For Superpowers changes how it feels to work with an agent, and that takes a session or two to get used to. If you are used to getting code in the first reply, the questions will feel slow. They are the point. The methodology trades a fast first draft for a plan you can trust and code that has been tested and reviewed, and for anything bigger than a toy, that trade wins. It also costs more tokens than letting an agent freestyle. Sub-agents per task, reviews between tasks, and tests before code all add up. The cheaper executing-plans mode implements every task inline in the current session with one fresh review of the whole branch at the end, instead of a sub-agent per task, if you want to save on that cost. Finally, Superpowers does not accept agent-written pull requests that fix issues filed by someone else unless a maintainer asked for one, and it does not generally accept contributions of new skills. Skill updates must work across all the coding agents it supports. Those rules keep the methodology consistent, but they mean the project grows carefully rather than fast. For users, that is a feature: the workflow you learn today is the workflow that will still be there tomorrow. FAQs What is Superpowers? A complete software development methodology for coding agents, built by Jesse Vincent and Prime Radiant. It is a set of composable skills that make your agent brainstorm before building, plan in bite-sized tasks, write tests first, and build with sub-agents and reviews. It is open source under MIT and had 309,623 stars in October 2026. How does Superpowers change the way my coding agent works? Instead of jumping straight into code, your agent first asks what you really want, plays its understanding back for correction, writes a design, breaks it into 2 to 5 minute tasks with exact files and verification steps, then builds with tests first and a review after each task. The skills trigger automatically and are mandatory workflows, not suggestions. Which coding agents does Superpowers work with? Claude Code, Cursor, Codex, GitHub Copilot CLI, Gemini CLI, and many more, including Antigravity, Devin, Factory Droid, Grok Build, Kimi Code, OpenCode, Pi, Qwen Code, Hermes Agent, and Muse. Install it separately for each harness you use. How do I install Superpowers in Claude Code? Run /plugin install superpowers@claude-plugins-official in Claude Code, or register the Superpowers marketplace with /plugin marketplace add obra/superpowers-marketplace and install from there. Start a fresh session and describe something to build; brainstorming should trigger before any code is written. Does Superpowers use test-driven development? Yes, true red-green TDD is a core rule. The agent writes a failing test, watches it fail, writes minimal code to pass, watches it pass, then refactors. Code written before tests gets deleted, and a verification skill checks work is actually complete before the agent declares success. What is subagent-driven development in Superpowers? After the plan is approved, fresh sub-agents work through each engineering task, with inspection and review after each one. A cheaper alternative, executing-plans, implements every task inline in the current session with one fresh review of the whole branch at the end. Is Superpowers free? Yes. It is open source under the MIT license. Prime Radiant offers commercial services for enterprises that want support or managed spending, but the methodology itself is free. What if Superpowers stops working in a long session? Start a fresh session, since the bootstrap that triggers skills is injected at session start and a compacted session can lose it. You can also ask your agent to figure out what went wrong with Superpowers in that session, and the diagnosing skill will report what happened with line-level evidence. --- Originally published at devpik.com. Repo: obra/superpowers.

How I spec out features with Superpowers How I spec out features with Superpowers I use Superpowers to work through features before handing them to a coding agent to implement. I want the agent to help me figure out what I'm asking for, including the decisions I haven't thought through yet. One feature I've been thinking about is letting a design doc contain additional documents. Take a Redis migration: you've got the spec and implementation plan, but now you want to rehearse it in staging and need a runbook. The initial request is pretty simple: let someone create that runbook inside the design doc. But if they follow the runbook, record a failed rehearsal, and then change the instructions, what happens to the record of that first attempt? Someone looking at it later needs to see the instructions that were actually used. That's already more specific than “let people add documents.” It's also something I'd want decided before an agent starts building it. Start with the problem For this feature, I'd give the agent the migration example and work through what the team needs to do: - Keep the staging runbook with the migration it belongs to. - Have a separate runbook for production. - Record what happened during each rehearsal. - Be able to look back at the instructions used for an earlier attempt. That brings up questions about ownership, permissions, and revisions. Can a runbook be shared across unrelated changes? Does it inherit the design doc's permissions? What happens to old references if someone renames or archives it? For the first version, I'd keep each supporting document attached to one change and use that change's permissions. Sharing a runbook across migrations and automatically executing its commands would be separate work. I want those decisions written down so the agent has a clear scope to work within. Where Superpowers helps Superpowers has a brainstorming skill that works through the request and possible approaches before moving into a design and implementation plan. It also tells the agent to inspect the project before asking questions the code or existing docs already answer. That last part matters to me. I don't want to point the agent at every relevant document or explain something it could find in the codebase. I want it to look first, then ask about what's still unresolved. With the runbook example, an agent could ask whether I want version history. I'd probably say yes, but that answer leaves a lot unspecified. A more useful question would be: > Someone follows revision 1 in staging and records a failed attempt. They change the instructions and try again using revision 2. What should someone reviewing the first attempt be able to see? That question puts me in the position of someone actually using the product. I can think about the engineer reviewing the failed rehearsal, what they're trying to understand, and what they'd need to see. Just asking “Do you want version history?” makes it easy to say yes without thinking through how anyone will use it. From that answer, we can get to a specific requirement: the first attempt has to keep a reference to revision 1. Editing the document can't change what that reference opens. Permissions need the same attention. Inheriting the parent's permissions should cover old revisions, search results, and access through the agent, as well as the page someone opens in their browser. These are the details I want to work through in the conversation. One thing I don't like One of my big gripes with using Superpowers is that I still have to tell the agent to do outside research. I mean searching the web, reading current documentation, and looking at how other people have approached the problem. I think an informed decision needs more than what the model learned during training. Reading my codebase gives it context about the project. I also want it to look for information outside the project that could change its recommendation. For the runbook example, I'd want it to investigate how other systems handle changes to a procedure after someone has used it. What tradeoffs did they make? What problems did they run into? That gives us more to consider when deciding how our version should work. I want the agent to recognize when that research would help and do it without me having to ask. Then show me the sources, explain how they apply to this feature, and make it clear which conclusions are its own. Having to remember to add “go research this” is a part of the workflow I'd like to improve. Keep the spec and implementation plan separate In my workflow, the spec describes what the change needs to accomplish. The implementation plan covers how to build it. Part of the spec for this feature could be: > A supporting document belongs to one design doc and inherits its access permissions. > > A recorded rehearsal can reference the exact document revision used. > > Editing, renaming, or archiving the document preserves that reference for authorized readers. > > Creating a runbook doesn't establish that the migration succeeded. The implementation plan can then work out how to store those references, which UI and agent operations need updating, and how to test them. The agent should also check what it can reuse. If the project already has revision history, I want the plan to explain how this feature fits into it. Introducing another versioning system would need a reason. Make the requirements testable I'd take the revision requirement and turn it into a scenario: 1. Create a staging runbook. 2. Record a failed rehearsal against revision 1. 3. Edit the runbook, producing revision 2. 4. Record another rehearsal against revision 2. 5. Open the first rehearsal's reference and confirm it still shows revision 1. An implementation could have a history page and still get this wrong. The scenario tells the agent exactly which behavior needs to hold up. The runbook describes the procedure. The rehearsal record says what happened when someone followed it. Both need to stay understandable after the instructions change. The same issue comes up elsewhere. If a pricing rule changes after an invoice is issued, you still need to know which rule produced that invoice. Thinking through a sequence like that helps me get more specific about what the feature needs to preserve. A starting prompt If you have Superpowers installed, you could adapt this: Use Superpowers to help me work through this feature before implementation. Problem: [Describe a specific situation and what someone needs to accomplish.] Look through the relevant code, docs, and previous decisions first. Tell me what already exists that affects this change. Identify decisions that would benefit from outside research and research them before recommending an approach. Look for current documentation and relevant examples of how others solved similar problems. Link the sources and explain how they affect your recommendation. Distinguish what the sources establish from your own conclusions. Ask about decisions we still need to make. Explain what the options would mean for the feature and recommend an approach. Make it clear when something is an assumption or your proposal. Write a spec covering the intended behavior, what's in and out of scope, unresolved questions, and concrete acceptance scenarios. Once we've agreed on the spec, create an implementation plan based on the codebase. Include how we'll check that the important behavior works. Give it an actual situation to work through. For a notification feature, for example, explain who needs to be notified, what happened, and what they need to do next. How are other people doing this? Superpowers is what I use, but I'd like to hear from people using other tools. Their docs describe a few approaches: | Tool | Documented workflow | | --- | --- | | GitHub Spec Kit | Takes a feature specification through a technical plan, tasks, and implementation, with project principles captured separately. | | OpenSpec | Keeps a proposal, requirements and scenarios, technical design, and tasks together for a change. The workflow includes archiving changes and updating specs. | | Kiro | Uses requirements, design, and task documents. Feature specs can start from requirements or from a technical design. | I'm also interested in people using their agent's planning mode and a Markdown file. Sometimes I just want to work through a problem without going through a full feature workflow. One thing I'm still thinking about is how to handle changes once implementation starts. The agent finds that an assumption was wrong, the plan changes, and the next session needs to know why. I keep a work log alongside the design doc to record what changed and why. How do you handle that in your setup? Do you update the spec, keep a separate record, or rely on the PR and conversation history? I'd also like to hear about a recent change where planning helped, or where it became more work than it was worth. Which tool were you using, and what did you end up keeping or skipping?

Spec first, tests first, then parallel agents: what I changed after months with obra/superpowers I write most of my code with Claude Code these days. For a long time my workflow was built on obra/superpowers, a plugin of skills that steers the agent through brainstorming, planning and test-driven development. It changed how I work, and it's still a great place to start. After a few months I had a list of things I wanted to work differently. That list became pasadena, a Claude Code plugin for spec-driven development. This post explains what it does and why. The problem: context that lives nowhere A spec says what you intend to build. Git history says what you merged. Neither says where the work stopped, what you already verified, or which file to open next. With agents it gets worse: every new session starts cold, and you spend the first ten minutes explaining everything again. Pasadena keeps that context in a journal file on the task branch, .pasadena/journal/<branch>.md. It has three sections: the goal; what's happening now (current phase, blocker, the next concrete step with a verification command); and an append-only timeline. Hooks restore the journal at session start, log new commits, and mark pauses and failures. Before merge, the useful summary moves into the PR description and the file is deleted. Built for worktrees Pasadena is designed around git worktrees: one task, one branch, one worktree. For architectural work, sheldon creates the worktree, the journal and a draft PR right away, so several tasks can run side by side without stepping on each other, and each one keeps its own journal on its own branch. The workflow /pasadena picks up from whatever stage the branch is at. Under the hood it hands off to one of nine skills: | Skill | Job | |---|---| | sheldon | Asks focused questions and writes an agreed spec | | penny | Builds 2–4 browser prototypes for a UI decision | | leonard | Splits the spec into tasks and parallel waves in plan mode | | wolowitz | Runs the waves, verifies and commits each task, reports on the PR | | amy | Failing tests first; proof from a fresh run before "done" | | leslie | Root cause in four phases before any fix | | bernie | End-to-end check in the browser, or a lasting Playwright test | | stuart | Screenshots at several widths and a report of layout defects | | raj | Starts, resumes, hands off and closes the journal | What I changed compared to superpowers 1. Plan mode for the plan. Once the spec is agreed, leonard writes the plan inside Claude Code's native plan mode. I review and approve it like any other plan, and nothing runs before that. 2. Codex as a helper, not a replacement. If the Codex CLI is installed and logged in, skills delegate three kinds of work to codex exec: bulk code lookups that come back as facts with file:line; an extra review of the whole branch; and plan tasks explicitly marked for Codex. Claude stays the orchestrator and only counts findings it has confirmed itself. The main motivation is saving Claude usage on routine work. 3. Spec-driven and test-driven in one pipeline. The spec sets the acceptance criteria. amy turns them into failing tests before any production code, then takes each step through its own failing test. Nothing counts as done without a fresh test run. 4. Parallel subagents without collisions. Every task in the plan declares the files it touches. leonard groups tasks whose file sets don't overlap into waves. wolowitz runs each wave in parallel, with a separate test writer and implementer per task, verifies it, commits every task, and reviews every third wave plus the whole branch at the end. 5. Work happens in the open, on a draft PR. The draft PR is opened at the start and carries the spec. While the work goes on, progress is commented on the PR as commits land, so the PR itself shows what was built, in what order and how it was verified, long before it's ready for review. Trying it /plugin marketplace add dmytro-rudenko/pasadena /plugin install pasadena@pasadena Requirements: bash, git, jq, and Node.js for browser prototypes. The plugin works on its own, but it works best together with the Codex CLI and microsoft/TokenEconomizer: Codex takes over part of the work, and TokenEconomizer helps spend fewer tokens. Wrapping up I use pasadena every day, and it covers everything I need in development. It keeps evolving. But that's one person's view, so I'm genuinely curious how it works for others. Any feedback is welcome, here in the comments or in issues on GitHub.

バイブコーディングで機能を継ぎ足していて見えてきたClaudeCodeの実装のクセ 「業務フローがよくわからないんだよね」という状態でも、チャラくバイブコーディングで自力でシステムが作れる時代が来ました。特にエンジニアは、アイディアをすぐに形にできます。普段からClaude CodeなどのAIエージェントに慣れ親しんでいれば、バイブコーディングのハードルは数ミリ程度まで下がっています。その結果、要件定義できるほどの解像度はない、できてからいろいろ意見はしたい、でも人の介在は最小限にしたい、という ガチのシステム屋さんに言ったら殴られそうな 要望を持つことも多いでしょう。 TL;DR 要件が固まらないままClaude Codeでバイブコーディングしたら、テストは通るのに画面ごとに数字や挙動が食い違った AIは依頼された画面の周りしか読まず、全体像や過去の経緯は勝手に把握してくれない 防ぐには、概念台帳・画面一覧・型とフックなどを人間が先に用意する必要がある 生成AI時代こそ、全体を設計して指揮するPMスキルが必須になる 何をどう作っていたか 開発対象 売上計画を立てられる社内システムです。月次の売上と在庫の見込み、売上計画、受注一覧、請求、入金、商品マスタ、顧客を扱います。 受注一覧 → 請求 → 実績 → 入金 → 現預金 → 粗利 といった流れでお金が流れる想定です。取引数や管理対象がまだ少ない段階で、事業計画を立てるためにシミュレーションをしたいという背景で作りました。 開発フロー 小規模なアプリ用に社内でテンプレートを作って、それをベースに開発していました。 工程 人 AI P1 要件整理 → Issue 一言で頼み、5つの質問に答える 5問を聞く → 要件メモ → Issueを起票 P2 実装 - 失敗するテストを書く → 最小の実装 → 整理 P3 品質ゲート → PR - format、lint、型、テスト、ビルド → PR作成 P4 レビュー → merge PRを読んでmerge （任意）AIレビュー P1の要件定義の5つの質問は、誰のためか、どの痛みか、なぜ今か、やらないこと、完成の定義です。superpowersのブレインストーミングを使っていて、質問の選択肢もAIが出してくれるので、適当に選びます。 管理していた文書 置き場 中身 00_process/ 開発フロー、矛盾を防ぐ進め方など 01_product/ デザインの指針、システム自体の細かい方針 02_architecture/adr/ ADR（設計判断とその理由の記録） 要件メモ/ 5問の答えから作ったメモ packages/shared/src の型とコメント 実質の仕様はここ（約1.8万行） 「要件メモ」の中にsuperpowersで作られる要件定義の資料はできあがります。ただ、今回のように「作りながら要件が固まる、変更がある」「最初に業務フロー全体像が見えていない」といった場合、途中で変更すると矛盾が生じます。AIが毎回このメモを更新して、唯一の正（SSOT）として活用してくれるわけではないからです。 大まかな欠点としては、 ドメイン知識や言葉が指示でぶれると、システムの中でもぶれる→概念台帳や用語集が必要 画面単位で改修をすることが多いので、既存実装の全体像を知らないまま重複して実装したり整合性が崩壊したりする→画面一覧などの一覧性のあるドキュメントが必要 といったところでしょうか。 この開発方法でぶちあたった課題 ざっと表にまとめてみました。 画面 何が食い違ったか 売上計画と受注一覧 途中で計画機能を追加した結果、受注の金額を計画でも一覧でも入力できるようになり、システム上で整合性がとれず画面ごとに数字がずれた。 請求 もともと税抜きで計算していたが税込にした。請求だけ税込への変換から漏れて、金額がずれた。AIは税金の扱いをコード上のコメントで管理していたらしい。 請求 入力を小数にするか整数にするか指定しなかったため、小数の入力ができなかった。（外注費はできるが売上はできない、といった挙動だった） 受注一覧 ページ構成を変更した時（入力が2か所あった箇所を、片方読み取り専用にして導線を整理した）、並び替え機能が消された。（ADRには並び替えの実装記録はあった） 画面全体 画面全体の書き出しと読み込みの機能について、実装している画面としていない画面があった（共通ルールになっておらず、途中で付け足した機能だったため）。さらに、実装が別々に書かれており整合性が崩れていた。 画面全体 畳める説明文の部品、記憶の読み書き、CSSのクラス名が重複して、結果的にデザインに統一性がない箇所がちらほらあった。共通部品の置き場までは定まっておらず、作る前に探す手順も定義されていなかった。 ADRで常に決定は残しておいたのですが、それでも矛盾が生じました。テストはそこそこ書いてくれていたのでCIは緑のまま、触ってみて「oh……」となりましたね。 あと細かいところだと、保守性の面で、特にフロントエンドはアーキテクチャを指定しても無視して1ファイルに書き始めたり、UI/UX的な考慮はされず横にたくさんスクロールが必要な表ができたりもしました。 AI（Claude Code）の実装のクセ AIの実装手順はこんな感じであろう 「売上計画に見込みの列を足して」と言った時の挙動は以下のようです（Claude自身が認識している方法ですが）。 必ず読む。CLAUDE.md、AGENTS.md、.claude/rules/ 言葉でgrepして対象のファイルを探す。「売上計画」「見込み」でgrepして、画面のファイルを開く importをたどる。開いたファイルから数段だけたどり、共通の計算関数を読む ADRや別の画面（受注一覧）は読まないことが多いようです。また、読む量が多すぎると、1で読んだはずの規則は華麗に無視してきます。こういうしくみなので、決めてから一気に組み上げるのには強く、運用ややりながら考えるのには弱い、という印象になるんですね。当たり前ですが、使っていると「なんでそんなこともわからんのだ……！」という気持ちになります。 ADRを読んで全体と経緯を把握してくれるわけではない 同僚が使っているClaude DesktopはADRを書いて設計したがり、指示をせずともADRを勝手に書いていました。が、Claude Codeで開発する時はあまりそんな気遣いは望めません。今回はAGENTS.mdに「ADRは必要になったら読む」くらいに書いていたのですが、常に読むわけではなく、読むとしてもgrepした結果の画面と関連がありそうなADRだけのようです。経緯を全部把握してくれる、みたいなことは期待しちゃいけない。さらに、後のADRが前の判断を覆している場合、そこまで読んでくれないこともあります。 IQというか処理能力が高くて知識も豊富だけど、めちゃくちゃおっちょこちょいでせっかちなジュニアの同僚、みたいな感じですね。 AIはシステムの全体像を勝手に把握するわけではない これも勝手にCLAUDE.mdとか何かしらからよしなに読むでしょ、と思っていました。人間でも最初に全体像がわかるようなドキュメントを読むじゃん。読まないんですよ。Claudeちゃんは究極のものぐさなので。 ユースケースなどの全体像がわからないと、業務の流れがわからないまま画面単位で追加の実装をはじめます。さらに、 セッションごとに記憶が切れる。前の画面で決めた流れは、書いていなければ残らない コードは画面単位（〜-board.tsx）に分かれている。読めるのは画面で、流れは読めない 受入条件は依頼された画面の中で閉じてしまい、AIが全体整合性をとらない といった特徴があります。E2Eテストをがんばればいいという話もありますが、結局人間が指示をちゃんと言語化しきらないといけないのは変わりません。 少なくとも要件定義で十分な情報をAIに言語化させて、全体像がわかる形でドキュメントにしておくことが必要です。 フロントのファイル分割ルールは無視するもの 部品の二重実装は、フロントのファイル分割を先に決めていれば防げる……そう思っていた時期が私にもありました。 Feature-Sliced Designを採用して、公開はindex.tsだけ、とルールを作ったのですが、何を思ったのかClaudeちゃんの実装は components/ に19ファイルが平置き（FSDガン無視） 受注一覧の画面が1ファイル2,110行、トップページが1,455行で、1画面を1ファイルで記載 CSSは1ファイルに2,382行で、分割する気0 features/ は2つだけで、共通部品を作れていない というポンコツっぷり。 AIは自らハーネスを作れない AIは自分の癖を把握して事前に防ぐ、といった気遣いはしてくれません。以下のような内容はハーネスとして準備しないと矛盾が起きる可能性があります。AIが苦手とする箇所ですね。 内容 対策 言葉の二重定義 正式な名前とコード上の名前を識別子つきで1か所で定義して、そこに無い識別子をhooksなどで検査する 値の表し方の揺れ 型で1本に縛る。空はnullだけ、単位は別の型にする、種別はunion型と一覧の1か所定義にする 計算の重複、および丸めなどの扱いの揺れ 計算をドメインの1関数に寄せて、画面のファイルからの計算関数のexportを禁止する。（可能ならDDDっぽい設計に寄せれば解決する） 数字の複数箇所での管理。入力口が複数になるとか、画面で再計算して結果がずれるなど 保存先に書き込める画面を1つに制限する。ほかの画面はドメインの結果を表示するだけにする 重複した実装をする 厳密なレイヤーおよび置き場の定義を行い、違反はlintで落とす 文書どうしが食い違う。台帳と実装がずれる 概念を足したら台帳に追記することをフックで強制する。文書の参照関係を検査する 特にアーキテクチャの面では、フロントエンドはかなり弱い印象があります。「自動で適切なものを選んで」ではなく、明示的に方針を指定したり、最初に雛形を用意してコンテキストに含めてあげたりする必要があるでしょう。 どう実装すればよかったか 社内でさくっと開発する、実際に使うまで細かいフローはわからん、といった状況の場合、 概念台帳および用語集 画面一覧と画面遷移 最低限の型とフック 受入条件（AC）をしっかり記載した実装用チケット 可能なら適度な粒度のアーキテクチャを実装したボイラープレート があるといいと思います。特に上の3つについて詳しくみます。 概念台帳と用語集 Spec駆動で開発しているつもりでも、途中で解像度が上がってくるとか、仕様自体が利用者の反応で変わるといったケースはあると思います。そうなるといつの間にかコードが正になり、背景や判断基準はADRを辿る必要がでてきます。また、指示する時の雰囲気で使う用語が変わると、AIはそれらを別物として扱ってしまいます。 とはいえ、がっつり要件定義ドキュメントを作るよりはAIに任せたい、といった場合、最低限「概念台帳」として、どこにどんな概念を置くかを管理しておくとよいです。 概念 正の置き場 表し方 単位 入力か導出か 受注の売上 受注一覧 小数を許す 税込、発生ベース 入力 ユースケースや業務フローから機能一覧を記載できると一番良いですが、結局その文章の整合性をとるには人間の時間と能力、そして経験が一定必要になってしまうので、社内でさくっと開発するには向かない感じがします。 画面一覧と画面遷移 画面と、その画面にある機能の一覧を1つのファイルで記載します。また、画面遷移もファイルにします。がっちがちの画面設計にする必要はなく、画面名、その画面でできること、入力か表示かあたりがあればいいでしょう。画面遷移もテキストの説明で十分です。 AIは依頼された画面のファイルから読み始めて、importを数段たどるだけなので、全体整合性をとるのが苦手です。毎回これらのファイルを読ませれば、全体の流れとしての不整合は防げます。ただし、これだけでは数値の丸め方など細かい規則まではフォローできないので、先の概念台帳と用語集のようなプチドメイン設計を併せてするとよいでしょう。 最低限の型とフック 文章で書いたルールはコードを読んでいるうちに忘れやがるので、静的検査で落とせるタイプのものはなるべくテストで落とすのも大事です。型がしっかりしていれば「ありえない値」などもある程度落とせるので、AIのゆるふわテストでもある程度は動くプロダクトになります。 全体を通して 今回の失敗から、「大規模開発だと生成AIの活用がうまくいかないよね」という話と同じ構造が見えました。大きく以下のような観点があると思います。 全体像が膨大で把握しきれない状況での、チームでの開発（人間のチームなのか、Claudeのセッションやサブエージェントなのかは違いますが） 要件定義フェーズをどう進めると抜け漏れが減るか、さらに抜け漏れがある前提でどう進めると事故が減るか 保守しやすいソフトウェアアーキテクチャなんかも大事ですが、いわゆる優秀なPMが今まで担ってきた内容が、生成AI時代に必須になっているように感じます。並列で動かす時はgit worktreeを使おうね、くらいは個人のエンジニアでもわかりそうですが、こういった全体の進め方についてはまだまだAIに全任せとはいかず、「知見と経験がある人がAIを使って全体をディレクションする」必要があるように思います。私は過激派なので、エンジニア経験のある人がみんなPMスキルを身につけるべき時代がくるぞとよく言います。できないPMの下でもやもやしていた実装経験のある人、今こそ輝くチャンスだぞ。あなたたちの力は絶対に未来に必要だ。 ちなみに 私はうぐいすソリューションズという会社を経営しています。弊社には役員が私含めて2人いて、それぞれ「要件定義からチーム開発をうまくいかせる鬼」と「技術全体像と適切なアーキテクチャ、そして状況整理マニア」という特性をもち、どちらも言語化に異常な執着があります。こういった知見をどんどん伝えてまともにPMができる人を増やそう、ある程度の再現性のある妥当で安心できるシステム開発をしよう、という組織なので、もし興味がある方がいればコーポレートサイトからぜひお問い合わせください。 uguisu-solutions.com カジュアルに話を聞いてみたい、PMやばい話を聞きたい、生成AI時代の愚痴が言いたいなど、なんでも結構です。同じ思いのある人で広く世の中をよくしていけたらな、という気持ちで、連絡をお待ちしています。

obra/superpowers and ovid/paad together work well for me flow is sp:brainstorm paad:pushback sp:write-plan paad:alignment sp:subagent-implementation - both have other useful skills in them but that is main flow

Test Results: we manually tried test-first + fresh verification on synthetic URL deduplication: 3 cases red before the fix, 4/4 green after. No native skill triggering, model comparison or measured time savings. mattpocock/skills obra/superpowers #OpenSource

فيه مهارة لكلود كود اسمها Superpowers، نزل لها تحديث جديد. صار يقرأ ملفات مشروعك ويفهم السياق قبل ما يبدأ يسألك، ويساعدك ترتب المتطلبات قبل التنفيذ. مفيدة للي يبني مشاريع كبيرة. obra/superpowers #ClaudeCode #AI

Copy any app, Claude does SEO, and the top 10 repos Link to Resource: thenextnewthing.ai - This video is sponsored by TripleTen, a professional education platform designed to help tech professionals and career changers move into higher-paying tech and AI roles. - Explore two TripleTen programs for experienced engineers: - AI Systems Engineering (22 weeks): tripleten.com - AI and Machine Learning (20 weeks): tripleten.com - Both are for engineers with 2+ years of recent, hands-on production experience and require a short skills audit before you can enroll. Both also include career coaching and mentorship to help you stand out in the job market. - TripleTen - github.com/tripleten-com - TripleTen Website - tripleten.com - Presented by Zapier: zapier.com - Zapier Next Gen Zaps - zapier.com This week, Andrew Warner is joined by Adam Brakhane to explore the GitHub repositories and AI tools that are making developers more productive. From running open-source apps in the cloud to giving AI coding agents security, design, and workflow superpowers, these projects show what's possible when developers build on top of the latest AI models. The episode starts with three bonus discoveries: a way to host open-source apps cheaply, Theo's collection of coding skills, and a project that makes iOS app development on Linux possible. Then Andrew and Adam count down the top 10 GitHub repositories of the week. They explore Cloudflare's security audit skill, Claude SEO, Impeccable, Superpowers, TradingAgents, Karpathy-inspired skills, design and accessibility tools, T3 Code, and Anthropic's Knowledge Work Plugins. Along the way, they discuss how open-source projects can become businesses, why skills need to be customized and kept up to date, how AI agents can be persuaded to follow instructions, and why the best coding tools are becoming more than just coding assistants. Repos Featured: Bonus Openers - Every App: every-app/every-app - T3 Stack: t3dotgg/tstack - Omarchy Apple Dev: joshuaswarren/omarchy-apple-dev Top 10 GitHub Repositories - Cloudflare Security Audit Skill: cloudflare/security-audit-skill - Claude SEO: AgriciDaniel/claude-seo - Impeccable: pbakaus/impeccable - Superpowers: obra/superpowers - TradingAgents: TauricResearch/TradingAgents - Andrej Karpathy Skills: multica-ai/andrej-karpathy-skills - Taste Skill: Leonxlnx/taste-skill - Interface Design Skills: jakubkrehel/skills - T3 Code: pingdotgg/t3code - Knowledge Work Plugins: anthropics/knowledge-work-plugins 00:00 - This Week's Best GitHub Repositories 00:27 - Every App: Host Open-Source Apps for $5 00:57 - Why Easy Deployment Matters for Open Source 02:15 - T3 Stack: Theo's AI Coding Skills 02:42 - Why You Should Customize Your AI Skills 03:43 - Omarchy Apple Dev: Build iOS Apps on Linux 04:29 - Why Apple Requires Mac Hardware for iOS Development 05:43 - #1 Cloudflare Security Audit Skill 07:00 - #2 Claude SEO: Run SEO Audits with Claude Code 09:31 - #3 Impeccable: Give AI Agents Better Design Taste 10:35 - Can Open-Source Design Tools Become Businesses? 12:40 - Presented by TripleTen 14:36 - #4 Superpowers: Give Coding Agents a Better Process 15:09 - Using Persuasion Principles to Improve AI Agents 17:11 - #5 TradingAgents: Simulated AI Trading 18:41 - #6 Andrej Karpathy Skills 20:31 - Why AI Agents Struggle to Follow Task Lists 21:43 - Presented by Zapier: Make AI Workflows Cheaper 22:38 - #7 Taste Skill: Give Your AI an Art Director 24:36 - #8 Interface Design Skills and Accessibility 26:20 - #9 T3 Code: Manage Multiple AI Coding Agents 27:20 - Run AI Agents Across Multiple Computers 30:46 - #10 Anthropic Knowledge Work Plugins 31:14 - What Should We Cover in First Stars? 31:45 - Adam's AI-Powered Business Context Layer 33:12 - Andrew's Automated AI Video Editing Workflow 34:18 - More GitHub Repositories to Explore 👉 Media/Sponsorship Inquiries: thenextnewthing.ai 👉 Join us: thenextnewthing.ai [de-DE] Jede App kopieren, Claude macht SEO und die Top 10 Repos Link zur Ressource: thenextnewthing.ai – Dieses Video wird von TripleTen gesponsert, einer professionellen Weiterbildungsplattform, die IT-Fachkräfte und Quereinsteiger beim Einstieg in besser bezahlte Positionen im Technologie- und KI-Bereich unterstützt. – Entdecken Sie zwei TripleTen-Programme für erfahrene Ingenieure: – KI-Systementwicklung (22 Wochen): tripleten.com – KI und Maschinelles Lernen (20 Wochen): tripleten.com – Beide Programme richten sich an Ingenieure mit mindestens zwei Jahren aktueller, praktischer Berufserfahrung. Vor der Anmeldung ist eine kurze Kompetenzanalyse erforderlich. Beide Programme beinhalten außerdem Karrierecoaching und Mentoring, um Ihnen zu helfen, sich auf dem Arbeitsmarkt hervorzuheben. - TripleTen - github.com/tripleten-com - TripleTen-Website - tripleten.com - Präsentiert von Zapier: zapier.com - Zapier Next Gen Zaps - zapier.com Diese Woche sprechen Andrew Warner und Adam Brakhane über GitHub-Repositories und KI-Tools, die die Produktivität von Entwicklern steigern. Von der Ausführung von Open-Source-Apps in der Cloud bis hin zur Ausstattung von KI-Programmieragenten mit Sicherheits-, Design- und Workflow-Funktionen – diese Projekte zeigen, was möglich ist, wenn Entwickler auf den neuesten KI-Modellen aufbauen. Die Folge beginnt mit drei Bonus-Entdeckungen: einer Möglichkeit, Open-Source-Apps kostengünstig zu hosten, Theos Sammlung von Programmierkenntnissen und einem Projekt, das die iOS-App-Entwicklung unter Linux ermöglicht. Anschließend präsentieren Andrew und Adam die Top 10 der GitHub-Repositories der Woche. Sie untersuchen Cloudflares Sicherheitsaudit-Funktion, Claude SEO, Impeccable, Superpowers, TradingAgents, von Karpathy inspirierte Funktionen, Design- und Barrierefreiheitstools, T3 Code und die Knowledge Work Plugins von Anthropic. Dabei erörtern sie, wie Open-Source-Projekte zu Unternehmen werden können, warum Fähigkeiten angepasst und aktuell gehalten werden müssen, wie KI-Agenten dazu gebracht werden können, Anweisungen zu befolgen, und warum die besten Codierungswerkzeuge mehr als nur Programmierassistenten sind. Empfohlene Repositories: Bonus-Opener - Every App: every-app/every-app - T3 Stack: t3dotgg/tstack - Omarchy Apple Dev: joshuaswarren/omarchy-apple-dev Top 10 GitHub-Repositories - Cloudflare Security Audit Skill: cloudflare/security-audit-skill - Claude SEO: AgriciDaniel/claude-seo - Impeccable: pbakaus/impeccable - Superpowers: obra/superpowers - TradingAgents: TauricResearch/TradingAgents - Andrej Karpathy Skills: multica-ai/andrej-karpathy-skills - Geschmacks-Skill: Leonxlnx/taste-skill - Interface-Design-Skills: jakubkrehel/skills - T3 Code: pingdotgg/t3code - Knowledge-Work-Plugins: anthropics/knowledge-work-plugins 00:00 – Die besten GitHub-Repositories der Woche 00:27 – Every App: Open-Source-Apps für 5 $ hosten 00:57 – Warum einfache Bereitstellung für Open Source wichtig ist 02:15 – T3 Stack: Theos KI-Programmierkenntnisse 02:42 – Warum Sie Ihre KI-Kenntnisse anpassen sollten 03:43 – Omarchy Apple Dev: iOS-Apps unter Linux entwickeln 04:29 – Warum Apple Mac-Hardware für iOS benötigt Entwicklung 05:43 – #1 Cloudflare-Sicherheitsaudit-Kompetenz 07:00 – #2 Claude SEO: SEO-Audits mit Claude Code durchführen 09:31 – #3 Tadellos: KI-Agenten ein besseres Gespür für Design vermitteln 10:35 – Können Open-Source-Designtools zu Unternehmen werden? 12:40 – Präsentiert von TripleTen 14:36 ​​– #4 Superkräfte: Optimieren Sie Ihre Programmieragenten 15:09 – KI-Agenten mithilfe von Überzeugungsprinzipien verbessern 17:11 – #5 TradingAgents: Simulierter KI-Handel 18:41 – #6 Andrej Karpathy Skills 20:31 – Warum KI-Agenten Schwierigkeiten haben, Aufgabenlisten abzuarbeiten 21:43 – Präsentiert von Zapier: KI-Workflows kostengünstiger gestalten 22:38 – #7 Geschmackssache: Geben Sie Ihrer KI einen Art Director 24:36 – #8 Interface-Design und Barrierefreiheit 26:20 – #9 T3 Code: Mehrere KI-Programmieragenten verwalten 27:20 – KI-Agenten auf mehreren Computern ausführen 30:46 – #10 Anthropic Knowledge Work Plugins 31:14 – Was sollten wir in First Stars behandeln? 31:45 – Adams KI-gestützte Business-Context-Ebene 33:12 – Andrews automatisierter KI-Video-Workflow 34:18 – Weitere GitHub-Repositories zum Entdecken 👉 Medien-/Sponsoringanfragen: thenextnewthing.ai 👉 Werden Sie Teil unserer Community: thenextnewthing.ai [es-US] Copia cualquier aplicación, Claude hace SEO y los 10 mejores repositorios Enlace al recurso: thenextnewthing.ai - Este video está patrocinado por TripleTen, una plataforma de formación profesional diseñada para ayudar a profesionales de la tecnología y personas que buscan un cambio de carrera a acceder a puestos mejor remunerados en tecnología e IA. - Descubre dos programas de TripleTen para ingenieros con experiencia: - Ingeniería de Sistemas de IA (22 semanas): tripleten.com - IA y Aprendizaje Automático (20 semanas): tripleten.com - Ambos programas están dirigidos a ingenieros con más de dos años de experiencia práctica reciente en producción y requieren una breve evaluación de habilidades antes de la inscripción. Ambos incluyen asesoramiento profesional y mentoría para ayudarte a destacar en el mercado laboral. - TripleTen - github.com/tripleten-com - Sitio web de TripleTen - tripleten.com - Presentado por Zapier: zapier.com - Zaps de nueva generación de Zapier - zapier.com Esta semana, Andrew Warner y Adam Brakhane exploran los repositorios de GitHub y las herramientas de IA que aumentan la productividad de los desarrolladores. Desde ejecutar aplicaciones de código abierto en la nube hasta dotar a los agentes de codificación de IA de capacidades avanzadas de seguridad, diseño y flujo de trabajo, estos proyectos demuestran el potencial de los desarrolladores que trabajan con los modelos de IA más recientes. El episodio comienza con tres descubrimientos adicionales: una forma económica de alojar aplicaciones de código abierto, la colección de habilidades de programación de Theo y un proyecto que permite el desarrollo de aplicaciones iOS en Linux. A continuación, Andrew y Adam presentan los 10 mejores repositorios de GitHub de la semana. Exploran la función de auditoría de seguridad de Cloudflare, Claude SEO, Impeccable, Superpowers, TradingAgents, las habilidades inspiradas en Karpathy, las herramientas de diseño y accesibilidad, T3 Code y los plugins de gestión del conocimiento de Anthropic. Además, analizan cómo los proyectos de código abierto pueden convertirse en negocios, por qué es necesario personalizar y actualizar las habilidades, cómo se puede persuadir a los agentes de IA para que sigan instrucciones y por qué las mejores herramientas de programación van más allá de ser simples asistentes. Repositorios destacados: Introducciones adicionales - Every App: every-app/every-app - T3 Stack: t3dotgg/tstack - Omarchy Apple Dev: joshuaswarren/omarchy-apple-dev Los 10 mejores repositorios de GitHub - Cloudflare Security Audit Skill: cloudflare/security-audit-skill - Claude SEO: AgriciDaniel/claude-seo - Impeccable: pbakaus/impeccable - Superpowers: obra/superpowers - TradingAgents: TauricResearch/TradingAgents - Andrej Karpathy Skills: multica-ai/andrej-karpathy-skills - Habilidad de Gusto: Leonxlnx/taste-skill - Habilidades de Diseño de Interfaz: jakubkrehel/skills - Código T3: pingdotgg/t3code - Plugins de Trabajo del Conocimiento: anthropics/knowledge-work-plugins 00:00 - Los mejores repositorios de GitHub de esta semana 00:27 - Every App: Aloja aplicaciones de código abierto por $5 00:57 - Por qué la implementación sencilla es importante para el código abierto 02:15 - T3 Stack: Las habilidades de programación de IA de Theo 02:42 - Por qué deberías personalizar tus habilidades de IA 03:43 - Omarchy Apple Dev: Crea aplicaciones iOS en Linux 04:29 - Por qué Apple requiere hardware Mac para iOS Desarrollo 05:43 - #1 Habilidad de auditoría de seguridad de Cloudflare 07:00 - #2 Claude SEO: Realiza auditorías SEO con Claude Code 09:31 - #3 Impecable: Mejora el diseño de los agentes de IA 10:35 - ¿Pueden las herramientas de diseño de código abierto convertirse en negocios? 12:40 - Presentado por TripleTen 14:36 ​​- #4 Superpoderes: Mejora el proceso de los agentes de codificación 15:09 - Uso de principios de persuasión para mejorar los agentes de IA 17:11 - #5 Agentes de comercio: Simulación de comercio con IA 18:41 - #6 Habilidades de Andrej Karpathy 20:31 - ¿Por qué los agentes de IA tienen dificultades para seguir las listas de tareas? 21:43 - Presentado por Zapier: Abarata los flujos de trabajo de IA 22:38 - #7 Habilidad de buen gusto: Asigna un director de arte a tu IA 24:36 - #8 Habilidades de diseño de interfaz y accesibilidad 26:20 - #9 Código T3: Gestiona múltiples agentes de codificación de IA 27:20 - Ejecuta agentes de IA en varios ordenadores 30:46 - #10 Plugins de trabajo de conocimiento antrópico 31:14 - ¿Qué deberíamos cubrir en First Stars? 3... [fr-FR] Copiez n'importe quelle application, Claude fait du SEO, et le top 10 des dépôts Lien vers la ressource : thenextnewthing.ai - Cette vidéo est sponsorisée par TripleTen, une plateforme de formation professionnelle conçue pour aider les professionnels de la tech et les personnes en reconversion à accéder à des postes mieux rémunérés dans les domaines de la tech et de l'IA. - Découvrez deux programmes TripleTen pour ingénieurs expérimentés : - Ingénierie des systèmes d'IA (22 semaines) : tripleten.com - IA et apprentissage automatique (20 semaines) : tripleten.com - Ces deux programmes s'adressent aux ingénieurs justifiant d'au moins deux ans d'expérience pratique récente en production. Une évaluation rapide de vos compétences est requise avant l'inscription. Ils incluent également un accompagnement et un mentorat pour vous aider à vous démarquer sur le marché du travail. - TripleTen - github.com/tripleten-com - Site web de TripleTen - tripleten.com - Présenté par Zapier : zapier.com - Zapiers nouvelle génération - zapier.com Cette semaine, Andrew Warner et Adam Brakhane explorent les dépôts GitHub et les outils d'IA qui améliorent la productivité des développeurs. De l'exécution d'applications open source dans le cloud à l'optimisation de la sécurité, du design et des flux de travail des agents de programmation IA, ces projets illustrent le potentiel des modèles d'IA les plus récents. L'épisode débute par trois découvertes bonus : une solution économique pour héberger des applications open source, les compétences de programmation de Theo et un projet permettant le développement d'applications iOS sous Linux. Andrew et Adam présentent ensuite leur top 10 des dépôts GitHub de la semaine. Ils explorent les fonctionnalités d'audit de sécurité de Cloudflare, Claude SEO, Impeccable, Superpowers, TradingAgents, les compétences inspirées de Karpathy, les outils de conception et d'accessibilité, T3 Code et les plugins Knowledge Work d'Anthropic. Au fil de leur exploration, ils abordent la transformation des projets open source en entreprises, la nécessité de personnaliser et de maintenir à jour les compétences, la manière d'inciter les agents d'IA à suivre des instructions et l'évolution des meilleurs outils de programmation, qui deviennent bien plus que de simples assistants de codage. Dépôts en vedette : Ouvertures bonus - Every App : every-app/every-app - T3 Stack : t3dotgg/tstack - Omarchy Apple Dev : joshuaswarren/omarchy-apple-dev Top 10 des dépôts GitHub - Cloudflare Security Audit Skill : cloudflare/security-audit-skill - Claude SEO : AgriciDaniel/claude-seo - Impeccable : pbakaus/impeccable - Superpowers : obra/superpowers - TradingAgents : TauricResearch/TradingAgents - Andrej Karpathy Skills : multica-ai/andrej-karpathy-skills - Compétences en perception gustative : Leonxlnx/taste-skill - Compétences en conception d'interfaces : jakubkrehel/skills - Code T3 : pingdotgg/t3code - Plugins pour le travail intellectuel : anthropics/knowledge-work-plugins 00:00 - Les meilleurs dépôts GitHub de la semaine 00:27 - Every App : Hébergez des applications open source pour 5 $ 00:57 - L'importance d'un déploiement facile pour l'open source 02:15 - T3 Stack : Les compétences de programmation IA de Theo 02:42 - Pourquoi personnaliser vos compétences en IA ? 03:43 - Omarchy Apple Dev : Créer des applications iOS sous Linux 04:29 - Pourquoi Apple exige du matériel Mac pour iOS Développement 05:43 - Compétence n° 1 : Audit de sécurité Cloudflare 07:00 - Claude SEO n° 2 : Réaliser des audits SEO avec Claude Code 09:31 - Impeccable n° 3 : Améliorer le sens du design des agents IA 10:35 - Les outils de conception open source peuvent-ils devenir rentables ? 12:40 - Présenté par TripleTen 14:36 ​​- N° 4 Superpouvoirs : Améliorer le processus des agents de codage 15:09 - Utiliser les principes de persuasion pour améliorer les agents d'IA 17:11 - N° 5 TradingAgents : Simulation de trading par IA 18:41 - N° 6 Compétences d'Andrej Karpathy 20:31 - Pourquoi les agents d'IA peinent à suivre les listes de tâches 21:43 - Présenté par Zapier : Réduire le coût des flux de travail d'IA 22:38 - N° 7 Sens du goût : Offrir un directeur artistique à votre IA 24:36 - N° 8 Compétences en conception d'interface et accessibilité 26:20 - N° 9 Code T3 : Gérer plusieurs agents de codage d'IA 27:20 - Exécuter des agents d'IA sur plusieurs ordinateurs 30:46 - N° 10 Plugins Anthropic Knowledge Work 31:14 - Que devrions-nous aborder dans First Stars ? 31:45 - Couche de contexte... [hi] किसी भी ऐप को कॉपी करें, Claude करेगा SEO, और टॉप 10 रेपोज़ संसाधन का लिंक: thenextnewthing.ai - यह वीडियो ट्रिपलटेन द्वारा प्रायोजित है, जो एक पेशेवर शिक्षा मंच है जिसे तकनीकी पेशेवरों और करियर बदलने वालों को उच्च वेतन वाली तकनीकी और एआई भूमिकाओं में आगे बढ़ने में मदद करने के लिए डिज़ाइन किया गया है। - अनुभवी इंजीनियरों के लिए ट्रिपलटेन के दो कार्यक्रमों के बारे में जानें: - एआई सिस्टम इंजीनियरिंग (22 सप्ताह): tripleten.com - एआई और मशीन लर्निंग (20 सप्ताह): tripleten.com - दोनों कार्यक्रम उन इंजीनियरों के लिए हैं जिनके पास हाल ही में 2+ वर्षों का व्यावहारिक उत्पादन अनुभव है और नामांकन से पहले एक संक्षिप्त कौशल ऑडिट आवश्यक है। दोनों कार्यक्रमों में करियर कोचिंग और मेंटरशिप भी शामिल है ताकि आप नौकरी बाजार में अलग पहचान बना सकें। - ट्रिपलटेन - github.com/tripleten-com - ट्रिपलटेन वेबसाइट - tripleten.com - ज़ैपियर द्वारा प्रस्तुत: zapier.com - ज़ैपियर नेक्स्ट जेन ज़ैप्स - zapier.com इस सप्ताह, एंड्रयू वार्नर के साथ एडम ब्राखाने GitHub रिपॉज़िटरी और AI टूल्स पर चर्चा करेंगे जो डेवलपर्स की उत्पादकता बढ़ा रहे हैं। क्लाउड में ओपन-सोर्स ऐप्स चलाने से लेकर AI कोडिंग एजेंट्स को सुरक्षा, डिज़ाइन और वर्कफ़्लो की अपार क्षमता प्रदान करने तक, ये प्रोजेक्ट दिखाते हैं कि जब डेवलपर्स नवीनतम AI मॉडल्स पर काम करते हैं तो क्या-क्या संभव है। एपिसोड की शुरुआत तीन अतिरिक्त खोजों से होती है: ओपन-सोर्स ऐप्स को सस्ते में होस्ट करने का एक तरीका, थियो के कोडिंग कौशल का संग्रह, और एक ऐसा प्रोजेक्ट जो लिनक्स पर iOS ऐप डेवलपमेंट को संभव बनाता है। इसके बाद एंड्रयू और एडम इस सप्ताह की शीर्ष 10 GitHub रिपॉज़िटरी की सूची प्रस्तुत करते हैं। वे क्लाउडफ्लेयर के सुरक्षा ऑडिट कौशल, क्लाउड एसईओ, इम्पेकेबल, सुपरपावर्स, ट्रेडिंग एजेंट्स, कार्पेथी से प्रेरित कौशल, डिज़ाइन और एक्सेसिबिलिटी टूल्स, टी3 कोड और एंथ्रोपिक के नॉलेज वर्क प्लगइन्स का अन्वेषण करते हैं। इस दौरान, वे चर्चा करते हैं कि कैसे ओपन-सोर्स प्रोजेक्ट व्यवसाय बन सकते हैं, कौशल को अनुकूलित और अद्यतन रखना क्यों आवश्यक है, एआई एजेंटों को निर्देशों का पालन करने के लिए कैसे प्रेरित किया जा सकता है, और क्यों सर्वश्रेष्ठ कोडिंग उपकरण केवल कोडिंग सहायकों से कहीं अधिक बन रहे हैं। प्रमुख रिपॉज़िटरी: बोनस ओपनर्स - एवरी ऐप: every-app/every-app - T3 स्टैक: t3dotgg/tstack - ओमार्ची एप्पल देव: joshuaswarren/omarchy-apple-dev शीर्ष 10 GitHub रिपॉज़िटरी - क्लाउडफ्लेयर सिक्योरिटी ऑडिट स्किल: cloudflare/security-audit-skill - क्लाउड एसईओ: AgriciDaniel/claude-seo - इम्पेकेबल: pbakaus/impeccable - सुपरपावर्स: obra/superpowers - ट्रेडिंग एजेंट्स: TauricResearch/TradingAgents - आंद्रेज कार्पेथी स्किल्स: multica-ai/andrej-karpathy-skills - स्वाद कौशल: Leonxlnx/taste-skill - इंटरफ़ेस डिज़ाइन कौशल: jakubkrehel/skills - T3 कोड: pingdotgg/t3code - नॉलेज वर्क प्लगइन्स: anthropics/knowledge-work-plugins 00:00 - इस सप्ताह के सर्वश्रेष्ठ GitHub रिपॉज़िटरी 00:27 - एवरी ऐप: ओपन-सोर्स ऐप्स को $5 में होस्ट करें 00:57 - ओपन सोर्स के लिए आसान डिप्लॉयमेंट क्यों महत्वपूर्ण है 02:15 - T3 स्टैक: थियो के AI कोडिंग कौशल 02:42 - आपको अपने AI कौशल को क्यों अनुकूलित करना चाहिए 03:43 - ओमार्ची एप्पल देव: लिनक्स पर iOS ऐप्स बनाएं 04:29 - एप्पल को इसकी आवश्यकता क्यों है iOS डेवलपमेंट के लिए मैक हार्डवेयर 05:43 - #1 क्लाउडफ्लेयर सुरक्षा ऑडिट कौशल 07:00 - #2 क्लाउड एसईओ: क्लाउड कोड के साथ एसईओ ऑडिट करें 09:31 - #3 इम्पेकेबल: एआई एजेंटों को बेहतर डिज़ाइन क्षमता प्रदान करें 10:35 - क्या ओपन-सोर्स डिज़ाइन टूल व्यवसाय बन सकते हैं? 12:40 - ट्रिपलटेन द्वारा प्रस्तुत 14:36 ​​- #4 सुपरपावर्स: कोडिंग एजेंट्स को बेहतर प्रक्रिया प्रदान करें 15:09 - एआई एजेंट्स को बेहतर बनाने के लिए अनुनय सिद्धांतों का उपयोग 17:11 - #5 ट्रेडिंग एजेंट्स: सिम्युलेटेड एआई ट्रेडिंग 18:41 - #6 आंद्रेज कार्पेथी स्किल्स 20:31 - एआई एजेंट्स को टास्क लिस्ट का पालन करने में कठिनाई क्यों होती है 21:43 - जैपियर द्वारा प्रस्तुत: एआई वर्कफ़्लो को सस्ता बनाएं 22:38 - #7 टेस्ट स्किल: अपने एआई को एक आर्ट डायरेक्टर दें 24:36 - #8 इंटरफ़ेस डिज़ाइन स्किल्स और एक्सेसिबिलिटी 26:20 - #9 टी3 कोड: कई एआई कोडिंग एजेंट्स को प्रबंधित करें 27:20 - कई कंप्यूटरों पर एआई एजेंट्स चलाएं 30:46 - #10 एंथ्रोपिक नॉलेज वर्क प्लगइन्स 31:14 - हमें फर्स्ट स्टार्स में क्या शामिल करना चाहिए? 31:45 - एडम का एआई-आधारित बिज़नेस कॉन्टेक्स्ट लेयर 33:12 - एंड्रयू का स्वचालित एआई वीडियो एडिटिंग वर्कफ़्लो 34:18 - एक्सप्लोर करने के लिए और भी GitHub रिपॉज़िटरीज़ 👉 मीडिया/स्पॉन्सरशिप पूछताछ: thenextnewthing.ai 👉 हमसे जुड़ें: thenextnewthing.ai [id] Salin aplikasi apa pun, Claude melakukan SEO, dan 10 repositori teratas Tautan ke Sumber Daya: thenextnewthing.ai - Video ini disponsori oleh TripleTen, platform pendidikan profesional yang dirancang untuk membantu para profesional teknologi dan mereka yang ingin beralih karier ke peran teknologi dan AI dengan gaji lebih tinggi. - Jelajahi dua program TripleTen untuk insinyur berpengalaman: - Rekayasa Sistem AI (22 minggu): tripleten.com - AI dan Pembelajaran Mesin (20 minggu): tripleten.com - Kedua program ini ditujukan untuk insinyur dengan pengalaman produksi langsung minimal 2 tahun dan memerlukan audit keterampilan singkat sebelum Anda dapat mendaftar. Kedua program ini juga mencakup pelatihan karier dan bimbingan untuk membantu Anda menonjol di pasar kerja. - TripleTen - github.com/tripleten-com - Situs Web TripleTen - tripleten.com - Dipersembahkan oleh Zapier: zapier.com - Zapier Next Gen Zaps - zapier.com Minggu ini, Andrew Warner ditemani oleh Adam Brakhane untuk menjelajahi repositori GitHub dan alat AI yang membuat pengembang lebih produktif. Mulai dari menjalankan aplikasi open-source di cloud hingga memberikan agen pengkodean AI kemampuan super dalam hal keamanan, desain, dan alur kerja, proyek-proyek ini menunjukkan apa yang mungkin terjadi ketika pengembang membangun di atas model AI terbaru. Episode dimulai dengan tiga penemuan bonus: cara untuk menghosting aplikasi open-source dengan murah, koleksi keterampilan pengkodean Theo, dan proyek yang memungkinkan pengembangan aplikasi iOS di Linux. Kemudian Andrew dan Adam menghitung mundur 10 repositori GitHub teratas minggu ini. Mereka mengeksplorasi kemampuan audit keamanan Cloudflare, Claude SEO, Impeccable, Superpowers, TradingAgents, kemampuan yang terinspirasi oleh Karpathy, alat desain dan aksesibilitas, T3 Code, dan Plugin Kerja Pengetahuan Anthropic. Sepanjang perjalanan, mereka membahas bagaimana proyek sumber terbuka dapat menjadi bisnis, mengapa kemampuan perlu disesuaikan dan diperbarui, bagaimana agen AI dapat dibujuk untuk mengikuti instruksi, dan mengapa alat pengkodean terbaik menjadi lebih dari sekadar asisten pengkodean. Repositori Unggulan: Pembuka Bonus - Every App: every-app/every-app - T3 Stack: t3dotgg/tstack - Omarchy Apple Dev: joshuaswarren/omarchy-apple-dev 10 Repositori GitHub Teratas - Cloudflare Security Audit Skill: cloudflare/security-audit-skill - Claude SEO: AgriciDaniel/claude-seo - Impeccable: pbakaus/impeccable - Superpowers: obra/superpowers - TradingAgents: TauricResearch/TradingAgents - Andrej Karpathy Skills: multica-ai/andrej-karpathy-skills - Taste Skill: Leonxlnx/taste-skill - Keterampilan Desain Antarmuka: jakubkrehel/skills - Kode T3: pingdotgg/t3code - Plugin Kerja Pengetahuan: anthropics/knowledge-work-plugins 00:00 - Repositori GitHub Terbaik Minggu Ini 00:27 - Every App: Host Aplikasi Open-Source seharga $5 00:57 - Mengapa Penyebaran yang Mudah Penting untuk Open Source 02:15 - T3 Stack: Keterampilan Pengkodean AI Theo 02:42 - Mengapa Anda Harus Menyesuaikan Keterampilan AI Anda 03:43 - Omarchy Apple Dev: Membangun Aplikasi iOS di Linux 04:29 - Mengapa Apple Membutuhkan Perangkat Keras Mac untuk Pengembangan iOS 05:43 - Keterampilan Audit Keamanan Cloudflare #1 07:00 - Claude SEO #2: Jalankan Audit SEO dengan Claude Kode 09:31 - #3 Sempurna: Berikan Agen AI Selera Desain yang Lebih Baik 10:35 - Bisakah Alat Desain Sumber Terbuka Menjadi Bisnis? 12:40 - Dipersembahkan oleh TripleTen 14:36 ​​- #4 Kekuatan Super: Berikan Agen Pengkodean Proses yang Lebih Baik 15:09 - Menggunakan Prinsip Persuasi untuk Meningkatkan Agen AI 17:11 - #5 TradingAgents: Simulasi Perdagangan AI 18:41 - #6 Keterampilan Andrej Karpathy 20:31 - Mengapa Agen AI Kesulitan Mengikuti Daftar Tugas 21:43 - Dipersembahkan oleh Zapier: Membuat Alur Kerja AI Lebih Murah 22:38 - #7 Keterampilan Selera: Berikan AI Anda Direktur Seni 24:36 - #8 Keterampilan Desain Antarmuka dan Aksesibilitas 26:20 - #9 Kode T3: Mengelola Beberapa Agen Pengkodean AI 27:20 - Menjalankan Agen AI di Beberapa Komputer 30:46 - #10 Plugin Kerja Pengetahuan Antropik 31:14 - Apa yang Harus Kita Bahas di Bintang Pertama? 31:45 - Lapisan Konteks Bisnis Bertenaga AI Adam 33:12 - Alur Kerja Pengeditan Video AI Otomatis Andrew 34:18 - Lebih Banyak Repositori GitHub untuk Dijelajahi 👉 Pertanyaan Media/Sponsor: thenextnewthing.ai 👉 Bergabunglah dengan kami: thenextnewthing.ai [it] Copia qualsiasi app, Claude fa SEO e le 10 migliori repository Link alla risorsa: thenextnewthing.ai - Questo video è sponsorizzato da TripleTen, una piattaforma di formazione professionale pensata per aiutare i professionisti del settore tecnologico e chi desidera cambiare carriera a trovare ruoli più remunerativi nel campo della tecnologia e dell'intelligenza artificiale. - Scopri due programmi TripleTen per ingegneri esperti: - Ingegneria dei sistemi di IA (22 settimane): tripleten.com - IA e apprendimento automatico (20 settimane): tripleten.com - Entrambi i programmi sono rivolti a ingegneri con almeno 2 anni di esperienza pratica recente in produzione e richiedono una breve valutazione delle competenze prima dell'iscrizione. Entrambi includono anche coaching e mentoring per aiutarti a distinguerti nel mercato del lavoro. - TripleTen - github.com/tripleten-com - Sito web di TripleTen - tripleten.com - Presentato da Zapier: zapier.com - Zapier Next Gen Zaps - zapier.com Questa settimana, Andrew Warner e Adam Brakhane esplorano i repository GitHub e gli strumenti di intelligenza artificiale che stanno rendendo gli sviluppatori più produttivi. Dall'esecuzione di app open source nel cloud al potenziamento di agenti di programmazione basati su IA con funzionalità avanzate di sicurezza, design e gestione dei flussi di lavoro, questi progetti dimostrano cosa è possibile realizzare quando gli sviluppatori si basano sui modelli di intelligenza artificiale più recenti. L'episodio inizia con tre scoperte bonus: un modo per ospitare app open source a basso costo, la raccolta di competenze di programmazione di Theo e un progetto che rende possibile lo sviluppo di app iOS su Linux. Infine, Andrew e Adam presentano la classifica dei 10 migliori repository GitHub della settimana. Esplorano la funzionalità di audit di sicurezza di Cloudflare, Claude SEO, Impeccable, Superpowers, TradingAgents, le competenze ispirate a Karpathy, gli strumenti di design e accessibilità, T3 Code e i plugin Knowledge Work di Anthropic. Nel corso della presentazione, discutono di come i progetti open source possano trasformarsi in attività commerciali, perché le competenze debbano essere personalizzate e costantemente aggiornate, come sia possibile persuadere gli agenti IA a seguire le istruzioni e perché i migliori strumenti di programmazione stiano diventando molto più che semplici assistenti alla programmazione. Repository in evidenza: Bonus Openers - Every App: every-app/every-app - T3 Stack: t3dotgg/tstack - Omarchy Apple Dev: joshuaswarren/omarchy-apple-dev I 10 migliori repository GitHub - Cloudflare Security Audit Skill: cloudflare/security-audit-skill - Claude SEO: AgriciDaniel/claude-seo - Impeccable: pbakaus/impeccable - Superpowers: obra/superpowers - TradingAgents: TauricResearch/TradingAgents - Andrej Karpathy Skills: Italiano: multica-ai/andrej-karpathy-skills - Competenza nel gusto: Leonxlnx/taste-skill - Competenze di progettazione dell'interfaccia: jakubkrehel/skills - Codice T3: pingdotgg/t3code - Plugin per il lavoro basato sulla conoscenza: anthropics/knowledge-work-plugins 00:00 - I migliori repository GitHub di questa settimana 00:27 - Every App: ospita app open source per 5 $ 00:57 - Perché la facilità di implementazione è importante per l'open source 02:15 - Stack T3: le competenze di programmazione AI di Theo 02:42 - Perché dovresti personalizzare le tue competenze AI 03:43 - Omarchy Apple Dev: crea app iOS su Linux 04:29 - Perché Apple richiede hardware Mac per lo sviluppo iOS 05:43 - #1 Competenza per l'audit di sicurezza di Cloudflare 07:00 - #2 Claude SEO: Eseguire audit SEO con Claude Code 09:31 - #3 Impeccable: Dare agli agenti IA un gusto migliore per il design 10:35 - Gli strumenti di design open source possono diventare aziende? 12:40 - Presentato da TripleTen 14:36 ​​- #4 Superpoteri: Migliorare il processo degli agenti di programmazione 15:09 - Utilizzare i principi della persuasione per migliorare gli agenti IA 17:11 - #5 TradingAgents: Simulazione di trading tramite IA 18:41 - #6 Competenze di Andrej Karpathy 20:31 - Perché gli agenti IA faticano a seguire le liste di attività 21:43 - Presentato da Zapier: Rendere i flussi di lavoro IA più economici 22:38 - #7 Abilità di gusto: Assegnare un direttore artistico alla tua IA 24:36 - #8 Competenze di progettazione dell'interfaccia e accessibilità 26:20 - #9 Codice T3: Gestire più agenti di programmazione IA 27:20 - Eseguire agenti IA su più computer 30:46 - #10 Plugin per il lavoro basato sull... [pt-BR] Copie qualquer aplicativo, Claude fazendo SEO e os 10 principais repositórios Link para o recurso: thenextnewthing.ai - Este vídeo é patrocinado pela TripleTen, uma plataforma de educação profissional criada para ajudar profissionais de tecnologia e pessoas em transição de carreira a ingressarem em cargos mais bem remunerados nas áreas de tecnologia e IA. - Explore dois programas da TripleTen para engenheiros experientes: - Engenharia de Sistemas de IA (22 semanas): tripleten.com - IA e Aprendizado de Máquina (20 semanas): tripleten.com - Ambos são destinados a engenheiros com mais de 2 anos de experiência prática recente em produção e exigem uma breve avaliação de habilidades antes da inscrição. Ambos também incluem coaching de carreira e mentoria para ajudar você a se destacar no mercado de trabalho. - TripleTen - github.com/tripleten-com - Site da TripleTen - tripleten.com - Apresentado por Zapier: zapier.com - Zaps de Próxima Geração do Zapier - zapier.com Nesta semana, Andrew Warner recebe Adam Brakhane para explorar os repositórios do GitHub e as ferramentas de IA que estão tornando os desenvolvedores mais produtivos. Desde a execução de aplicativos de código aberto na nuvem até a atribuição de superpoderes de segurança, design e fluxo de trabalho a agentes de codificação de IA, esses projetos mostram o que é possível quando os desenvolvedores criam com base nos modelos de IA mais recentes. O episódio começa com três descobertas bônus: uma maneira de hospedar aplicativos de código aberto a baixo custo, a coleção de habilidades de programação de Theo e um projeto que torna possível o desenvolvimento de aplicativos iOS no Linux. Em seguida, Andrew e Adam apresentam a lista dos 10 principais repositórios do GitHub da semana. Eles exploram a habilidade de auditoria de segurança da Cloudflare, o Claude SEO, o Impeccable, o Superpowers, o TradingAgents, habilidades inspiradas em Karpathy, ferramentas de design e acessibilidade, o T3 Code e os plugins de trabalho intelectual da Anthropic. Ao longo do caminho, discutem como projetos de código aberto podem se tornar negócios, por que as habilidades precisam ser personalizadas e mantidas atualizadas, como agentes de IA podem ser persuadidos a seguir instruções e por que as melhores ferramentas de programação estão se tornando mais do que apenas assistentes de codificação. Repositórios em destaque: Aberturas bônus - Every App: every-app/every-app - T3 Stack: t3dotgg/tstack - Omarchy Apple Dev: joshuaswarren/omarchy-apple-dev Top 10 Repositórios do GitHub - Cloudflare Security Audit Skill: cloudflare/security-audit-skill - Claude SEO: AgriciDaniel/claude-seo - Impeccable: pbakaus/impeccable - Superpowers: obra/superpowers - TradingAgents: TauricResearch/TradingAgents - Andrej Karpathy Skills: multica-ai/andrej-karpathy-skills - Habilidade de Paladar: Leonxlnx/taste-skill - Habilidades de Design de Interface: jakubkrehel/skills - Código T3: pingdotgg/t3code - Plugins de Trabalho Intelectual: anthropics/knowledge-work-plugins 00:00 - Os Melhores Repositórios do GitHub desta Semana 00:27 - Every App: Hospede Aplicativos de Código Aberto por US$ 5 00:57 - Por que a Facilidade de Implantação é Importante para o Código Aberto 02:15 - T3 Stack: As Habilidades de Programação de IA do Theo 02:42 - Por que Você Deve Personalizar Suas Habilidades de IA 03:43 - Omarchy Apple Dev: Crie Aplicativos iOS no Linux 04:29 - Por que a Apple Exige Hardware Mac para iOS Desenvolvimento 05:43 - #1 Habilidade de Auditoria de Segurança do Cloudflare 07:00 - #2 Claude SEO: Execute Auditorias de SEO com o Claude Code 09:31 - #3 Impecável: Dê aos Agentes de IA um Melhor Senso de Design 10:35 - Ferramentas de Design de Código Aberto Podem se Tornar Negócios? 12:40 - Apresentado pela TripleTen 14:36 ​​- #4 Superpoderes: Dê aos Agentes de Codificação um Processo Melhor 15:09 - Usando Princípios de Persuasão para Aprimorar Agentes de IA 17:11 - #5 TradingAgents: Simulação de Negociação por IA 18:41 - #6 Habilidades de Andrej Karpathy 20:31 - Por que os Agentes de IA Têm Dificuldade em Seguir Listas de Tarefas 21:43 - Apresentado pela Zapier: Torne os Fluxos de Trabalho de IA Mais Baratos 22:38 - #7 Habilidade de Paladar: Dê à Sua IA um Diretor de Arte 24:36 - #8 Habilidades de Design de Interface e Acessibilidade 26:20 - #9 Código T3: Gerencie Vários Agentes de Codificação de IA 27:20 - Execute Agentes de IA em Vários Computadores 30:46 - #10 Plugins de Trabalho de Conhecimento Antrópico 31:14 - O que Devemos Abordar no First Stars? 31:45 - Camada d...

1- Superpowers : Superpowers, kodlama ajanlarınız için tam bir yazılım geliştirme metodolojisidir; bir dizi kompozisyon becerisi ve ajanınızın bunları kullanmasını sağlayan bazı ilk talimatlar üzerine kuruludur. obra/superpowers