Mentions
Conversations linking to GitHub, newest first. Times are UTC.
Mentions by source
X
Bluesky
Hacker News
V2EX
Qiita
DEV
Mastodon
Latest posts
[ANN] console: making OCaml CLI tools look good! I'm happy to announce the release of console 0.9.0, a set of widgets for building nice-looking command-line tools. The main design constraint I tried to follow is that the output should look good both for a human at an interactive terminal and for scripts run in CI or other non-interactive loops (like LLM agents). Live rows are redrawn in place and everything useful the program printed is still there when it ends; piped output tries to be relatively clean.  You can use it with: opam install console console-eio The widgets in console are pure; console-eio is the driver that runs animations. There are a few limitation (including limited Windows support) but you can already use it to encode quite complex UX. For instance, here a few imitations of package manager frontends using console (recorded using k7): | | | | --- | --- | |  |  | |  |  | You can read more (including how it compares with notty, progress and printbox, here): gazagnaire.org. Feedback is very welcome before 1.0, on the issue tracker or here.
Hermes Onyx 1.0 - BOTS!! preview.redd.it preview.redd.it Marre de devoir rester planté devant mon navigateur sur PC pour suivre ou lancer des tâches avec mes agents IA, du coup j'ai développé Hermes Onyx ces derniers mois. L'idée, c'est une app Android native, 100 % open source et sans aucune télémétrie, reliée à votre serveur local ou hébergé (Ollama, vLLM, Hermes Agent). Vos clés et vos données restent chez vous. Ce qu'elle permet : - Déclencher des tâches et envoyer des commandes vocales depuis votre montre (Wear OS). - Interagir en mains libres et écouter des résumés contextuels en voiture (Android Auto). - Suivre les logs d'exécution et gérer les autorisations des outils sur téléphone ou tablette. Le code est dispo ici :JemaOS/hermes-onyx Le projet est encore tout récent, donc je cherche du monde pour tester, casser le truc et remonter les bugs. Si vous faites tourner des modèles en local, vous voudriez lui faire faire quoi en priorité depuis votre montre ou en voiture ?
jwangkun/dsh-superagent 已经封装好了插件,大家可以安装使用了 t.co
打字打到手腕疼的,试试直接张嘴。 OpenCode 团队出的 Hex,按住快捷键说话,松手字就粘到光标那儿。 听写在本机跑,第一次下个模型,之后声音不出你电脑。 免费,MIT 开源,Mac/Win/Linux 都有。 我写文案最卡第一句,说出来可能比憋出来快。 你们用语音输入吗? anomalyco/hex t.co
well if u wanna try opentui-split: npm i opentui-split github: viralcodex/opentui-split
Precio del m² en Medellín por estrato: apartamentos frente a casas (venta, octubre 2026) En Medellín, el m² de un apartamento en venta va de 3,81 millones de pesos en estrato 1 a 8,65 millones de pesos en estrato 6. En estrato 4, el m² de una casa es 27,2 % más… DOI: 10.5281/zenodo.23518933
광 디스크에 이미지 그리는 프로그램 X에서 奶昔🥤 님 : "全能的光盘盘面可见图像生成与刻录工具 JiaFeiMiao-K-Cat/DiscImageStudio x.com" / X x.com GitHub - JiaFeiMiao-K-Cat/DiscImageStudio · GitHub JiaFeiMiao-K-Cat/DiscImageStudio
@stevierothy @crackdown OMG I'm working on a Crackdown 2 #recomp just working on performance and cross-platform ImpetuousDanny/Crackdown2-Recomp
riseup-oss/mcp לכל הפרטים על ה-MCP שלנו
I built a DSL that compiles into .pcap files I'm building FrameAssembly because simulating network traffic is a pain in the ass. Whenever I need specific packets (like SNMP traps) to test monitoring tools, writing Scapy scripts or googling pcaps sucks, and recording live traffic means setting up test gear and filtering out a bunch of random background noise. FrameAssembly just takes a writing simple code, describing the packets and spits out a .pcap or .pcapng file. Doesn't need root or raw sockets since it only writes the file, and when I actually want the traffic on an interface I run the .pcap through tcpreplay. GitHub: SteveWinschel/FrameAssembly Crates.io: crates.io
Structured AI output in Flutter: Gemini, OpenAI, Claude Short answer: all three APIs can return JSON that matches a schema you send. Each wants the schema in its own dialect: Gemini takes upper-case type names, OpenAI's strict mode needs every property required and additionalProperties: false, and Claude needs additionalProperties: false on every object. Write the schema once and convert it per vendor, or use gemini_openai_claude, which does the conversion and returns the decoded JSON. If your app reads a receipt, tags a note or fills a form from a photo, you want an object back, not prose you have to parse. This guide shows how to get that from all three vendors with one schema, and how to handle the failures that look like success. One request, three dialects Here is the same small schema the way you would write it: { "type": "object", "properties": { "title": { "type": "string" }, "mood": { "type": "string", "enum": ["calm", "busy"] } }, "required": ["title"] } And what each vendor actually needs: | | Gemini | OpenAI (strict) | Claude | |---|---|---|---| | Where the schema goes | generationConfig.responseSchema | response_format.json_schema with strict: true | output_config.format | | Type names | STRING, OBJECT | string, object | string, object | | additionalProperties: false | not needed | on every object | on every object | | Optional fields | as written | not allowed: list all in required, make the optional ones nullable | as written | | Optional enum | as written | must also list null | as written | | Images | inline_data parts after the text | image_url with a data: URL | base64 image blocks before the text | Without OpenAI's strict: true, the schema is only a hint and the reply may not match it. With it, an optional property makes the request fail with a 400. That is why mood above becomes "type": ["string", "null"] with null added to its enum before it goes to OpenAI. The important design choice is where these conversions live. If you write the Claude conversion inside your Claude code path, the other two keep working, your tests against the default provider pass, and the feature is quietly broken for everyone who switched. Keep the conversions next to the schema so one change reaches all three. The code import 'package:gemini_openai_claude/gemini_openai_claude.dart'; final spec = PromptSpec( instruction: 'Suggest a short title and a mood for this note: "Ship the release before Friday."', schema: { 'type': 'object', 'properties': { 'title': {'type': 'string'}, 'mood': {'type': 'string', 'enum': ['calm', 'busy']}, }, 'required': ['title'], }, ); final AiAdapter ai = switch (provider) { 'gemini' => GeminiProvider(), 'openai' => OpenAiProvider(), _ => ClaudeProvider(), }; final json = await ai.complete(spec: spec, apiKey: key, model: model); // {"title": "Friday release", "mood": "busy"} The package is plain Dart with one dependency (http). It calls the REST APIs directly, so no vendor SDK update can break it. For a photo, add images: [PromptImage(mime: 'image/jpeg', base64: base64Encode(bytes))] to the spec. Failures that arrive as HTTP 200 Checking the status code is not enough. All three vendors can return 200 with no usable answer: - Safety block or refusal. Gemini returns no content, with a blockReason or a finishReason such as SAFETY or RECITATION. OpenAI fills a refusal field or ends with content_filter. Claude returns stop_reason: "refusal". - Cut off at the token limit. Gemini says MAX_TOKENS, OpenAI finish_reason: "length", Claude stop_reason: "max_tokens". The JSON looks almost valid and then stops mid-token. - Empty answer. A 200 with an empty candidate list. If you parse these as normal replies, your app reports "the API may have changed", which sends you to read code instead of retrying. The adapters turn each case into a ProviderException with its own kind: declined, truncated or emptyResult. The HTTP errors also deserve separate handling: | Status | Usually means | What to tell the user | |---|---|---| | 401, or 400 mentioning the key | Bad key | Check the key | | 403 on Gemini | API not enabled on the project, or region not supported | Not a key problem | | 404 | Model name retired or misspelled | Choose a current model | | 429 | Rate limit or billing | Wait, or switch provider | | 5xx, 529 on Claude | Vendor overloaded | Try again shortly | Model names are a setting Vendors retire model names roughly once a year. If the name is a constant in your app, the feature stops working the day it is retired and stays broken until you ship an update. Make the model a setting or a remote config value, and treat a 404 as "choose a current model", not as a crash. Where the API key lives Anything compiled into an app can be extracted, so do not ship your own API key in the binary. Two patterns are safe: 1. Users bring their own key. They paste it into settings, you store it in the Keychain or Keystore, and the app calls the vendor directly. Good for tools aimed at developers and power users. 2. Your server holds the key. The app calls your backend, and the backend calls the vendor. This is the only option when you pay for the requests. Because the package is plain Dart, the same adapters can run on a Dart server. Testing without a network Every adapter accepts an http.Client, so tests use MockClient and never call a vendor. DemoProvider answers on the device from an offline answer you attach to the spec, which is useful for demo builds and screenshots: final spec = PromptSpec( instruction: '...', schema: schema, demo: () => {'title': 'Friday release', 'mood': 'busy'}, ); await const DemoProvider().complete(spec: spec, apiKey: ''); FAQ How do I get structured JSON output from Gemini, OpenAI and Claude with the same schema? Write one JSON Schema and convert it per vendor: upper-case types for Gemini, strict mode with every property required for OpenAI, and additionalProperties: false on every object for Claude. The gemini_openai_claude package does this for you. Why does OpenAI reject my schema with a 400? In strict mode every property must be listed in required, every object needs additionalProperties: false, and an optional field must be nullable, including null in its enum. Why does Gemini return 403 with a valid API key? The Generative Language API is not enabled on the key's project, or the region is not supported. It is a project setting, not the key. Is it safe to put my OpenAI or Gemini API key in a Flutter app? No. Let users bring their own key, or call the vendor from your own server. Do I need the vendor SDKs? No. All three have REST APIs, and a thin HTTP client is easier to keep working across SDK releases. --- gemini_openai_claude is open source on GitHub. It comes from the AI apps in our Flutter app templates. Originally published at coderqh.com.
Source: DeepSeek's repo and its philosophy doc. DeepSeek labels it a research prototype, not an official product. deepseek-ai/dsh-superagent-prototype
Precio del m² en Manizales por estrato: apartamentos frente a casas (venta, octubre 2026) En Manizales, el m² de un apartamento en venta va de 4,66 millones de pesos en estrato 3 a 6,40 millones de pesos en estrato 6. En estrato 3, el m² de una casa es 33,7 %… DOI: 10.5281/zenodo.23521668
LordOS is a free Debian-based Linux distro with its own lightweight desktop, custom Wayland compositor, and privacy-focused native apps. It runs on PCs and Raspberry Pi 5 and stays open source. link; lordbasex/LordOS t.co
VideoDL(videofetch)是一款纯 Python 实现的轻量视频下载器,能够解析 B 站、抖音、快手、小红书、油管等一大批平台。 作者已经注明,它仅供学习使用、禁止商用;遇到付费内容,请先开通正版会员,尊重版权。 🔗CharlesPikachu/videodl t.co
Precio del m² en La Ceja por estrato: apartamentos frente a casas (venta, octubre 2026) En La Ceja, el m² de un apartamento en venta va de 5,10 millones de pesos en estrato 3 a 7,72 millones de pesos en estrato 5. En estrato 5, el m² de una casa es 22,7 % más… DOI: 10.5281/zenodo.23522379
Concrete test: identical q7/JSON/collector wording, but a token origin is flagged and public status data stays quiet on SEC-001/011. Four synthetic offline cases, reports and commands: github.com/fayzkk889/MCPSense/tree/main/docs/laun…
DS4Windows fork update: unified UI, faster profile editing, DS3 pressure support and more (5.0.60) Hi everyone, Following on from my previous post, I wanted to share what’s changed in my DS4Windows fork since 5.0.46. 5.0.60 is now available, bringing a substantial interface refresh, faster profile editing, more flexible profile management and several community-reported fixes. 🌐 Explore the website ⬇️ Download the latest release Thank you again to the original DS4Windows contributors for the work this builds on, and to everyone sharing suggestions, screenshots, profiles and logs. Quite a few of these improvements came directly from those conversations. Here’s an overview: Interface and usability - Standardised dropdowns, sliders, circular expanders and popup menus throughout the app, with consistent hover, pressed, selected, disabled and keyboard-focus states. - Improved consistency across all six themes, including the tray menu and its submenus. Checked menu items now align properly with unchecked items. - Added more compact headers, tidier spacing and better alignment throughout Overview, Controllers, Settings and Profile Editor. - Refreshed the footer with service status, controller count, expandable activity details and separate Hotkeys/About buttons, while keeping its compact height. - Improved tooltip behaviour and added an optional Show tooltips setting, enabled by default. - Refreshed the button-mapping dialog and Special Actions editor, including readable themed tabs, clearer controls and more consistent layouts. Profiles and faster editing - Added remembered Tiles, List and Details views. Tiles remains the default; Details shows the emulated controller and last-modified time. - Profile badges now use the first letter of the profile name. Choose a custom fill colour or use your own PNG/ICO image instead. Icons travel with the profile through duplication, import/export, backups and migration. - Rename and Duplicate now use matching inline panels, with clearer validation and protection against replacing an existing profile. - Added Settings & profiles controls for choosing the save location, opening the profiles folder, and backing up or restoring settings and profiles together. - Installed copies can optionally migrate settings from the program folder to AppData, with verified copies, backups and the original files retained. - The storage section stays collapsed by default, remembers your preference and automatically expands when the selected folder is read-only or requires attention. - Profile Editor is faster and smoother to open and close. Settings pages load when first needed, hidden controls avoid unnecessary discovery work, and status polling no longer stalls the interface behind busy audio configuration. Special Actions - Added optional triple-tap macros, configurable hold/tap timing, and individual names and enable switches for gestures. - Improved action importing with backups and protection against overwriting existing actions. - Fixed enabled actions becoming unchecked after saving when their names contained slashes. - Fixed overlapping Multi Actions causing a triple tap to also trigger matching single/double actions. - Added a shortcut back to Settings that offers to save or discard profile edits first. DualShock 3 pressure support - Added pressure-sensitive mappings for DS3 face buttons, D-pad and L1/R1 when assigned to analogue triggers or stick directions. - Partial presses and releases now retain their analogue values. Ordinary digital button, keyboard and macro assignments keep their existing behaviour. - This requires DsHidMini’s raw SXS input mode; its DS4-compatible mode does not expose those additional pressure channels. Controller information and audio - Added Controller info buttons to Overview and Controllers, separating information about the physical controller from the emulated device. - View available identity, battery, firmware and factory information, with selectable values, Refresh and Copy info. Fields depend on what each controller exposes. - Corrected malformed emulated DualShock 4 factory dates and audited the other emulated metadata formats. Emulated firmware dates are clearly distinguished from the DS4Windows version. - Virtual audio devices now have clear names such as DS4Windows Controller 1 Audio and DS4Windows Controller 1 Microphone, making them easier to distinguish from physical devices. - Improved supported Sony USB microphone forwarding, including profile volume, noise suppression and mute handling, without requiring a separate virtual audio cable. - Fixed DualShock 4 Bluetooth microphone activation when Windows opens the virtual microphone, including the audio companion used with Xbox output. Fixes and reliability improvements - Controller startup: Resolved the reported intermittent failure to reach Ready through paired DS4Windows/VIIPER communication fixes. Special thanks to cns00 for sticking with the lengthy investigation and providing repeated diagnostic captures. - Installer readiness: Fixed a separate timing race in the final health check that could incorrectly request repair even though setup had succeeded. - Profile switching: Restored two-finger swiping after switching profiles with the DualSense mute button, and improved notification requests after successful switches. Profile-change notifications are shown while the app is minimised to the tray. - Preventing double input: Added Keep controllers hidden between reconnects, enabled by default when using HidHide. This reduces the window in which games can discover the physical controller before its replacement virtual controller is ready. Stop/exit releases DS4Windows-owned retained entries. - Saving permissions: Installed copies default to AppData, read-only locations are clearly identified, and optional migration provides a safer alternative to running everything as administrator. - Startup warnings: Brief, expected USB haptics setup waits no longer immediately produce an alarming Needs attention status. Persistent waits and genuine failures remain visible. - Layout and navigation: Fixed unwanted initial focus, Auto Profiles opening partially scrolled down, dropdown hover causing scrolling, clipped controls and several light/dark theme discrepancies. Installer, update storage and documentation - The installer now reports HidHide and FakerInput status and versions, alongside DS4Windows, VIIPER and USB/IP. HidHide is selected by default only when missing; FakerInput remains optional. - Uninstall clearly explains that it removes DS4Windows and its managed VIIPER copy while preserving shared drivers, profiles and settings. Shortcuts help users find profiles or manage shared drivers separately. - Added configurable update storage management. Keep one previous version per component by default, choose another retention policy, inspect occupied/reclaimable space and preview cleanup. - Updated the documentation with current screenshots, clearer setup instructions and guidance matching the refreshed interface. - Expanded automated coverage for profile preservation, multiple-controller isolation, feedback handling and installer behaviour. - 5.0.60 includes VIIPER 0.1.17. USB/IP remains at 0.9.8.1, so no USB/IP driver version change is required. In progress Generic DirectInput output for older games remains in development separately. Community testing has been encouraging, and the branch now uses a 16-button layout with further mapping work planned. It is still awaiting PID approval and is not included in the stable release yet. I’m also continuing to investigate hardware-specific feedback reports. Additional diagnostics help distinguish missing game feedback from problems further along the controller connection, without assuming every symptom has the same cause. Thanks especially to cns00, Jobima1st, 1hezio, eVenent, Masterload1975, lpv11, pinn73, jusseppe88 and Darkii-0 for their suggestions, reports and testing. Links - Repository and overview - Full changelog - Troubleshooting - Report an issue - Discussions Please back up your profiles and close games, DS4Windows and VIIPER before updating. Use the complete package with its matching backend. Windows releases remain unsigned; checksums and build records accompany the downloads. If you try it, I’d love to hear which controller and connection you’re using, how it works for your setup, and what could be improved next.
Introducing @dhh’s SolidQueue Can run even with SQLite rails/solid_queue
Our config said automatic reasoning effort. Two launch paths dropped it. The no-mistakes review caught both while we built ReddiDotAgents to run Claude, Codex and Antigravity as one fleet. linkedin.com nissan/ReddiDotAgents #CodingAgents
Nothing new here for us, nor for several other accounts here that are familiar with the techniques; what we see is simply information that is already public and published online. jakejarvis/awesome-shodan-queries shodan.io t.co
Precio del m² en Jamundí por estrato: apartamentos frente a casas (venta, octubre 2026) En Jamundí, el m² de un apartamento en venta va de 3,20 millones de pesos en estrato 3 a 4,22 millones de pesos en estrato 5. En estrato 3, el m² de una casa es 19,6 % más… DOI: 10.5281/zenodo.23522327
Gaussian SplattingのPLY、中身が何だったか分からなくなりませんか?僕はよくなります。 なのでWindowsのエクスプローラーでPLYファイルのサムネイルを出すツール「SplatThumb」を作りました。アングルも設定できます。 無料・オープンソースです VoxelKei/SplatThumb #GaussianSplatting #3DGS t.co
I built GateKeepr — a local-first tool for systematic review screening with evidence-verified AI I've been working on a literature review and wanted a better way to handle full-text screening, so I built GateKeepr. It lets you: - Screen PDFs against your own inclusion/exclusion criteria. - Use optional Gemini/OpenAI analysis with quotations verified against the actual PDF. - Highlight evidence, track decisions, and export results. - Run everything locally, without needing a hosted service. AI only assists — the final screening decision is yours. It also works completely without AI. It's free to use, and the code is on GitHub. I'd love feedback from anyone doing systematic or literature reviews! GitHub: faraz-riaz/gatekeepr
Precio del m² en Itagüí por estrato: apartamentos frente a casas (venta, octubre 2026) En Itagüí, el m² de un apartamento en venta va de 4,82 millones de pesos en estrato 2 a 8,08 millones de pesos en estrato 5. En estrato 2, el m² de una casa es 25,2 % más… DOI: 10.5281/zenodo.23521973
Answer: 0 numbers, 0 proof, no stakes, buzzwords like "leverage synergies", and 4 of its 6 sentences start with "We". copy-gate: sales 35/100, motive 0/100, gate 25/100. Free and open source (MIT), zero deps, offline: sjgant80-hub/copy-gate As an agent skill: skillhq.dev The draft was written for this demo. Same input, same output hash on 2 runs.
Precio del m² en Ibagué por estrato: apartamentos frente a casas (venta, octubre 2026) En Ibagué, el m² de un apartamento en venta va de 3,49 millones de pesos en estrato 2 a 5,00 millones de pesos en estrato 5. En estrato 2, el m² de una casa es 39,1 % más… DOI: 10.5281/zenodo.23522188
DIY Nissan Leaf Dashboard — Arduino + OLED I'd like to share another small project for the Nissan Leaf — a custom auxiliary dashboard based on an Arduino and an OLED display. The goal was simple: to have convenient access to the main vehicle and battery parameters without overloading the factory dashboard or interfering with its operation. What's inside? The project is built around: - Arduino; - MCP2515 CAN controller (500 kbps, 8 MHz crystal); - SH1106 OLED display, 128×64; - Screen navigation using the factory gear selector. The dashboard receives data directly from the vehicle's CAN bus. What does it display? Several screens are available, each showing different information: - SOC and range — battery charge level and estimated driving range. - Power — current power consumption and regenerative braking. - Driving range — additional information for estimating the remaining range. - Battery — voltage, current, and basic battery parameters. - Telemetry — battery temperature, remaining capacity in Ah, cell voltage imbalance, minimum and maximum cell voltage. - Diagnostics — additional technical information. I also implemented support for custom CAN telemetry from my CAN bridge. This makes it possible to display battery parameters that are not available in the vehicle's standard CAN messages. Controls Screen navigation is handled using the factory gear selector, so there is no need to install additional buttons. The selected screen is saved in EEPROM. After restarting the vehicle, the display returns to the previously selected screen. Temperature filtering Special attention was given to battery temperature readings. Sometimes incorrect values can briefly appear on the CAN bus. To prevent sudden temperature jumps on the display, I implemented filtering that averages several measurements and checks suspicious changes. Source code The project is open source. The source code and connection details are available on GitHub: 👉 Arduino Nissan Leaf Dashboard — GitHub The project was developed using open-source libraries, components, and work from other developers. The original authors retain their credit, and all components are used in accordance with their original licenses. This is not a commercial product, but my own development for the Nissan Leaf, which I decided to share with the community. Feedback, suggestions, and ideas for further development are welcome!
QQQ-Anchored Quantitative Trading Strategy Suite 👇 (kemomi/qqq-anchored-quant) t.co
مشروع ترفيهي مفتوح المصدر يخلي Claude يدق على جوالك اذا يحتاج إجابة منك او انه خلص من المهمة AliCreating/claude-phone-arabic
@gigitlm github.com/makecindy/cindy/blob/main/README.zh-CN… 最近火的Cindy
Precio del m² en Floridablanca por estrato: apartamentos frente a casas (venta, octubre 2026) En Floridablanca, el m² de un apartamento en venta va de 4,17 millones de pesos en estrato 3 a 7,22 millones de pesos en estrato 6. En estrato 3, el m² de una casa es… DOI: 10.5281/zenodo.23522248
bonus: AI4Finance-Foundation/FinRL-Trading: AI4Finance-Foundation/FinRL-Trading FinRL’in yeni AI-native tarafı ML + reinforcement learning + risk overlay + backtest + execution’ı modüler şekilde birleştirmeye çalışıyor AI agent / RL ile trading deneyi yapmak isteyen biri için kurcalamaya değer ama burada özellikle dikkat: modelin karmaşıklığı edge olduğu anlamına gelmiyor basit baseline’ı geçemiyorsa 15 agent kullanmanın pek anlamı yok
New PR. mpv player now switches theme based on your @omarchy theme. @dhh @hancore_linux omarchy#15467 t.co
stefan-jansen/machine-learning-for-trading: stefan-jansen/machine-learning-for-trading bu tek bir trading botundan çok açık kaynak quant/ML laboratuvarı gibi data > feature > model > backtest > costs > risk > deployment zincirini uçtan uca ele alıyor özellikle ilgimi çeken taraf: lookahead bias walk-forward validation transaction costs overfitting AI agents RAG + financial research gibi çoğu “ai trading bot” demosunun atladığı şeyleri merkeze koyması
Precio del m² en El Retiro por estrato: apartamentos frente a casas (venta, octubre 2026) En El Retiro, el m² de un apartamento en venta va de 6,50 millones de pesos en estrato 3 a 10,15 millones de pesos en estrato 6. En estrato 5, el m² de una casa es 17,5 %… DOI: 10.5281/zenodo.23522003
The 2026 B2B SaaS SEO Architecture: Engineering Information Gain in the AI Search Era The traditional B2B SaaS SEO playbook of publishing thirty 2,000-word glossaries and waiting for pipeline to materialize is officially dead. With generative search models (Google AI Overviews, Perplexity, and ChatGPT search) synthesizing top-of-funnel answers directly on the search engine results page, generic SaaS content is bypassed before a human ever clicks. To build an organic customer acquisition channel that drives recurring revenue in 2026, engineering and marketing teams must transition from vanity keyword volume to structured information gain and entity authority. Here is the architectural blueprint we deploy at Windel Solutions for enterprise B2B SaaS platforms. --- 1. The Death of Keyword Density: Building Entity-Driven Topic Hubs Search engines no longer parse pages as bags of disconnected keywords. Modern algorithms evaluate topical authority through vector embeddings and knowledge graph entities. When a B2B SaaS targets high-intent search queries like enterprise search architecture or data pipeline governance, competing pages often have nearly identical copy. To stand out, your content architecture must execute three structural rules: 1. Tight Parent-Child Hierarchies: Avoid loose blog posts scattered across random subdirectories. Establish dedicated pillar hubs (for example, comprehensive guides such as B2B SaaS SEO Strategy) linked bidirectionally to hyper-focused sub-modules. 2. Contextual JSON-LD Schemas: Implement TechArticle, SoftwareApplication, and Service schema graphs. Search crawlers use structured data to verify company legitimacy and distinguish your brand entity from generic aggregator blogs. 3. Cross-Entity Verification: Explicitly map your organization schema to verified industry registries, including your Crunchbase profile, Clutch agency listings, and open-source GitHub repositories. --- 2. Information Gain Engineering: Why Search Bots Reward Original Data Google's information gain patents prioritize content that contributes novel data, unique code benchmarks, or contrarian findings not present in existing index documents. If your article merely summarizes the top three ranking Google results, your page will be relegated to the secondary crawl queue ("Crawled - currently not indexed"). To capture first-page positions: - Lead with Benchmarks: Publish real-world telemetry data (for instance: server latency metrics, database query bottlenecks, or conversion rate shift analysis). - Incorporate Interactive Tooling: Provide lightweight CLI diagnostic tools or open-source scripts directly within your technical write-ups. We maintain public audit utilities like the Windel Solutions Core Web Vitals Audit Toolkit to give engineers verifiable diagnostic capabilities. - Answer Edge Cases: Address technical trade-offs, architecture limitations, and rollback plans. B2B buyers search for how systems behave when scaling under load, not basic definitions. --- 3. Core Web Vitals and Crawl Budget Efficiency Topical authority is useless if your infrastructure prevents search bots from efficiently crawling your site. Enterprise SaaS websites frequently deploy complex single-page application (SPA) frameworks or client-side rendering libraries that block crawler rendering threads. Critical Performance Benchmarks: - Time to First Byte (TTFB): Must remain below 200ms. High server response times burn search engine crawl budget, causing new product feature pages to sit undiscovered for weeks. - Interaction to Next Paint (INP): Heavy analytics scripts, live chat widgets, and multi-tier tracking pixels freeze the main browser thread. Purging third-party script bloat directly accelerates user interaction and search rankings. Teams looking to systematically profile and eliminate render-blocking bottlenecks can reference our frameworks on B2B Site Speed Optimization Services and perform an Enterprise Technical SEO Audit. --- 4. The 90-Day Enterprise Organic Growth Roadmap If you are scaling a B2B SaaS platform from single-digit impressions to dominant Page 1 rankings: 1. Days 1 to 30: Technical Remediation: Fix redirect chains, eliminate orphan 404s, and compress main-thread JavaScript execution below 2.5 seconds. 2. Days 31 to 60: Information Gain Ingestion: Revise high-impression, low-click URLs with proprietary data tables, code snippets, and structured FAQ schemas. 3. Days 61 to 90: Authority Syndication: Leverage developer communities, digital PR citations, and open-source contributions to build natural referring domain velocity. Organic search remains the highest-margin acquisition channel in B2B SaaS, provided you treat SEO as an engineering discipline rather than a copywriting exercise. --- Published by Mustafa Saad, Founder & Lead Technical SEO Architect at Windel Solutions. Connect with our technical team to audit your enterprise search performance.
Precio del m² en Chía por estrato: apartamentos frente a casas (venta, octubre 2026) En Chía, el m² de un apartamento en venta va de 5,00 millones de pesos en estrato 2 a 9,93 millones de pesos en estrato 6. En estrato 6, el m² de una casa es 31,5 % más barato… DOI: 10.5281/zenodo.23521502
Of course I made a Firmware. Hello everyone, of course I made my own firmware. At first I wanted to recreate my old OP-Z, because I sold it (stupid me). It took a month to build and test it internally, and today is release day for the alpha version of "JAN3000". JanSchulten/JAN3000-M-Wave-Firmware jan3000-manual.vercel.app You can find the whole project on GitHub, with an installer and an online manual. Features 1. 4 tracks, 17 sound engines: virtual analog, 6-operator FM, wavetable, granular, physical modelling, organ, samples, a 303-style bass line and two drum engines 2. Built for shortcuts and muscle memory: hold a button and the keys and knobs change their job 3. 64-step sequencer per track with chords, chance, ratchets, nudge and parameter locks 4. Live recording including knob moves, with undo and redo 5. 16 scenes per project, switched in time and copied while playing 6. Song mode that chains patterns into a loop you can change while it plays 7. KEY track: a chord progression that moves all patterns along with it 8. 100 chord sets in many genres, chord keys and an arpeggiator 9. Punch-in FX in two key maps, recorded onto two dedicated FX tracks, plus insert, chorus, delay or granular cloud, reverb and slicer per track 10. MIDI over USB and TRS, MIDI learn, USB audio to your computer and a sync pulse out Future features - Using the second CPU core - Project backups - Teenage Engineering Pocket Operator sync (right now sync only works with other instruments like Behringer or Korg Volcas)
Rate Limiting in Rails 8.1: Protect Your APIs Without Rack::Attack --- title: Rate Limiting in Rails 8.1: Protect Your APIs Without Rack::Attack published: true date: 2026-11-02 10:41:02 UTC tags: rubyonrails,programming,ratelimiting,ruby canonical_url: medium.com --- Have you ever wondered what happens when someone sends hundreds of requests to your login endpoint within a few seconds? Without rate limiting, your application could face unnecessary database queries, excessive server load, brute-force login attempts, or API abuse. Rate limiting helps control how frequently a client can access an endpoint within a specified time window. The good news is that Rails provides a built-in rate_limit method, so you don't always need an additional gem to implement basic rate limiting. Rails 8.1 also introduces a useful improvement: the scope: option, which allows multiple controllers to share the same rate-limit counter. Let’s understand how it works with practical examples. 1. What Is Rate Limiting? Rate limiting is a technique used to restrict the number of requests a client can make within a specific period. For example: - Allow 5 login attempts within 3 minutes. - Allow 100 API requests per minute. - Allow 10 password reset requests within 15 minutes. When a client exceeds the configured limit, the application rejects additional requests. Usually, the server responds with: HTTP 429 Too Many Requests This helps protect application resources and reduces abuse. 2. Rails Built-in Rate Limiting Rails provides the rate_limit class method in Action Controller. You can configure it directly inside a controller without installing an additional gem. Let’s consider a login endpoint. # app/controllers/sessions_controller.rb class SessionsController < ApplicationController rate_limit to: 5, within: 3.minutes, only: :create def create # Login authentication logic end end Let’s understand the configuration: Option Meaning to: 5 Maximum 5 requests within: 3.minutes Within a 3-minute window only: :create Apply the limit only to the create action By default, Rails identifies requests using the client’s IP address. When the limit is exceeded, Rails raises ActionController::TooManyRequests, which Action Dispatch handles by returning HTTP 429. 3. Customize the Rate Limit Response The default response might not be sufficient for an API application. You can customize the response using the with: option. class SessionsController < ApplicationController rate_limit to: 5, within: 3.minutes, only: :create, with: -> { render json: { error: "Too many login attempts. Please try again later." }, status: :too_many_requests } def create # Login authentication logic end end Now, when a client exceeds the configured limit, the API returns a JSON error response. Example: { "error": "Too many login attempts. Please try again later." } With HTTP status: 429 Too Many Requests This is particularly useful when building APIs consumed by React, mobile applications, or other external clients. 4. Rate Limit Based on User Instead of IP By default, Rails applies rate limiting based on the request IP address. But what if you want to apply limits per authenticated user? You can customize the identifier using the by: option. class Api::BaseController < ApplicationController rate_limit to: 100, within: 1.minute, by: -> { current_user&.id || request.remote_ip } end Here: - Authenticated users are identified by their user ID. - Unauthenticated requests are identified by their IP address. - Each identifier gets its own rate-limit counter. This is useful when you want to provide individual API limits to authenticated users. Important: If your application runs behind a proxy or load balancer, configure trusted proxy handling correctly. Otherwise, IP-based limits may not identify clients as expected. 5. What’s New in Rails 8.1? The scope: Option This is one of the most useful improvements to the built-in rate-limiting functionality. Before Rails 8.1, rate-limit counters were scoped to individual controller paths by default. Consider an API application with two controllers: Api::PostsController Api::UsersController Suppose both controllers have the following configuration: rate_limit to: 100, within: 1.minute By default, these controllers maintain separate counters. This means a client could make: - 100 requests to the posts endpoint. - 100 requests to the users endpoint. That’s 200 requests in total. What if you want both controllers to share one common limit? Rails 8.1 introduces the scope: option for this purpose. Example: Shared rate limit across controllers Create a common API controller: # app/controllers/api_controller.rb class ApiController < ApplicationController rate_limit to: 100, within: 1.minute, scope: "public_api" end Now inherit from it: class Api::PostsController < ApiController def index render json: Post.all end end class Api::UsersController < ApiController def index render json: User.all end end Both controllers share the same rate-limit scope. If a client makes 60 requests to the posts endpoint, only 40 requests remain for the users endpoint during the same rate-limit window. The shared counter uses the same scope and client identifier. Why is this useful? Imagine an application exposing 20 public API endpoints. Instead of allowing 100 requests per endpoint, you can enforce a common limit across the entire API. This gives you better control over overall API usage. 6. Multiple Rate Limits on the Same Controller Sometimes one rate limit isn’t enough. For example, you might want: - A short-term limit of 3 requests every 2 seconds. - A longer-term limit of 10 requests every 5 minutes. Rails supports multiple rate limits on the same controller. Use the name: option to distinguish them. class SessionsController < ApplicationController rate_limit to: 3, within: 2.seconds, name: "short_term" rate_limit to: 10, within: 5.minutes, name: "long_term" def create # Login logic end end Each limit maintains its own counter. This is useful when you want to prevent both sudden bursts of traffic and repeated requests over a longer period. 7. Where Does Rails Store Rate-Limit Counters? An important question is: How does Rails remember how many requests a client has already made? Rails uses an ActiveSupport::Cache store to maintain rate-limit counters. By default, it uses the controller cache store, which falls back to the application’s configured cache store. For a production application running on multiple servers, using a shared Redis cache can be useful. Example: class ApiController < ApplicationController RATE_LIMIT_STORE = ActiveSupport::Cache::RedisCacheStore.new( url: ENV["REDIS_URL"] ) rate_limit to: 100, within: 1.minute, store: RATE_LIMIT_STORE end Now the rate-limit counters are stored in Redis. Why does this matter? Imagine your application runs on three application servers: Load Balancer | ------------------- | | | Server 1 Server 2 Server 3 If each server maintains its own independent counters, a client might bypass the intended limit by sending requests to different servers. Using a shared Redis cache allows all three servers to access the same counters. Make sure Redis is configured with appropriate availability, memory, and eviction settings for your production workload. 8. Rails Rate Limiting vs Rack::Attack Before Rails introduced built-in rate limiting, developers commonly used gems such as Rack::Attack. Does Rails’ built-in functionality mean you no longer need Rack::Attack? Not necessarily. !cdn-images-1.medium.com For straightforward controller-level rate limiting, Rails’ built-in functionality is often sufficient. If you need advanced request filtering, blocklists, safelists, or more complex throttling rules, Rack::Attack is still worth considering. 9. When Should You Use Rails Built-in Rate Limiting? Here are some practical use cases: Login endpoints Prevent repeated login attempts that could indicate brute-force activity. Password reset endpoints Avoid excessive password reset requests. Public APIs Control how frequently clients can access your API. Registration endpoints Reduce automated account creation attempts. Expensive operations Protect endpoints that perform costly database queries or external API calls. Remember that rate limiting is only one layer of application protection. You should also implement authentication, authorization, input validation, monitoring, and appropriate infrastructure-level protections. Conclusion Rails makes basic rate limiting easier by providing the rate_limit method directly in Action Controller. With Rails 8.1, the new scope: option makes it possible to share rate-limit counters across multiple controllers. The key takeaways: - Use rate_limit to restrict requests within a time window. - Use by: to customize client identification. - Use with: to customize the response. - Use scope: to share limits across controllers. - Use store: to configure a shared cache such as Redis. - Use name: when defining multiple limits on one controller. You don’t always need an additional gem to protect your Rails endpoints from excessive requests. Sometimes, the framework already provides the functionality you need. Have you used Rails’ built-in rate limiting, or do you prefer Rack::Attack? Share your experience in the comments! References - Rails 8.1 Action Controller Rate Limiting API - Rails 8.1 Action Pack Changelog - Rack::Attack GitHub Repository
Precio del m² en Cartagena por estrato: apartamentos frente a casas (venta, octubre 2026) En Cartagena, el m² de un apartamento en venta va de 4,57 millones de pesos en estrato 3 a 10,06 millones de pesos en estrato 6. En estrato 3, el m² de una casa es 37,7 %… DOI: 10.5281/zenodo.23521365
Test your own agent in 3 commands: pip install invaris-agentsec agentsec init agentsec test invarislabs/invaris-agentsec
Next.js OG images without the @vercel/og limits next/og (the ImageResponse API, built on @vercel/og) is the default way to make Open Graph images in Next.js, and for a title on a gradient it is excellent. It is not a browser, though: it converts JSX to SVG with Satori and rasterises it with Resvg. When your card needs CSS grid, your real stylesheet, woff2 fonts or more than 500 KB of assets, the fix is not to fight Satori — it is to render the card with real Chrome and put the resulting URL in generateMetadata. This guide lists the limits precisely, shows when they bite, and gives a drop-in replacement that works on the Node.js and Edge runtimes. The limits, from the Next.js docs The ImageResponse reference states them plainly: - Only flexbox and a subset of CSS. "Advanced layouts (e.g. display: grid) will not work." The full list of supported properties is in Satori's README. - 500 KB bundle limit, counting your JSX, CSS, fonts, images and other assets together. - Fonts: ttf, otf and woff only — no woff2, which is what most font services and next/font hand you. - Default output is 1200×630 PNG, which is the right size for Open Graph. In practice people hit these as: a two-column card that needs grid; a brand font that only ships as woff2 (convert it, or it silently falls back); two weights of a CJK font that blow past 500 KB on their own; a logo as SVG with filters Resvg does not support; and the general cost of re-implementing a design in inline styles because the site's CSS is not available. A note if you read Satori's own README: recent standalone Satori releases added CSS grid and WOFF2, but the @vercel/og that Next.js bundles (1.0.x in Next 16, built on an older Satori) does not have them — the Next.js limits above are the ones that apply to next/og. None of this is a bug. Satori trades completeness for running anywhere in milliseconds. If your card fits inside it, keep it. Option A: keep next/og, but stay inside the lines If the limits are only fonts and size, fix those and stay with ImageResponse: // app/blog/[slug]/opengraph-image.tsx import { ImageResponse } from 'next/og'; import { readFile } from 'node:fs/promises'; import { join } from 'node:path'; export const size = { width: 1200, height: 630 }; export const contentType = 'image/png'; // ttf/otf parse fastest; subset the font to Latin to stay far below 500 KB. const inter = readFile(join(process.cwd(), 'assets/Inter-Bold-latin.ttf')); export default async function Image({ params }: { params: Promise<{ slug: string }> }) { const post = await getPost((await params).slug); return new ImageResponse( ( <div style={{ display: 'flex', flexDirection: 'column', justifyContent: 'space-between', width: '100%', height: '100%', padding: 72, background: '#0b0a09', color: '#fff' }}> <div style={{ fontSize: 64, fontWeight: 700, lineHeight: 1.1 }}>{post.title}</div> <div style={{ fontSize: 28, opacity: 0.7 }}>yourblog.com</div> </div> ), { ...size, fonts: [{ name: 'Inter', data: await inter, weight: 700, style: 'normal' }] }, ); } Subsetting fonts (for example with pyftsubset from fonttools) is the single most effective fix for the 500 KB limit. Option B: real Chrome, as a URL in generateMetadata When you need grid, your stylesheet, woff2, emoji fonts or complex scripts, render the card from HTML and CSS in a real browser. Running Chrome inside a serverless function is possible (a trimmed Chromium build such as @sparticuz/chromium with puppeteer-core), but cold starts of several seconds and memory limits make it fragile — and a link-preview crawler will not wait. The lighter option is a signed image URL: you compute it in generateMetadata (no network call, no API key on the page), and the image service renders it on the first request and serves the cached file afterwards. Below is Shareglint's URL format (disclosure: I build Shareglint) with its built-in blog-og template — swap in your own HTML template's slug to use your own design. Web Crypto makes the same code run on Node.js and Edge: // lib/og.ts const enc = (v: string) => encodeURIComponent(v).replace(/[!'()*]/g, (c) => '%' + c.charCodeAt(0).toString(16).toUpperCase()); export async function ogImageUrl(template: string, fields: Record<string, string>) { const path = `/i/${process.env.OG_ACCOUNT_ID}/${template}.png`; const query = Object.keys(fields).sort().map((k) => `${enc(k)}=${enc(fields[k])}`).join('&'); const key = await crypto.subtle.importKey( 'raw', new TextEncoder().encode(process.env.OG_SIGNING_SECRET!), { name: 'HMAC', hash: 'SHA-256' }, false, ['sign'], ); const mac = await crypto.subtle.sign('HMAC', key, new TextEncoder().encode(`${path}?${query}`)); const s = Array.from(new Uint8Array(mac), (b) => b.toString(16).padStart(2, '0')).join('').slice(0, 32); return `https://shareglint.com${path}?${query}&s=${s}`; } // app/blog/[slug]/page.tsx import type { Metadata } from 'next'; import { ogImageUrl } from '@/lib/og'; export async function generateMetadata({ params }: { params: Promise<{ slug: string }> }): Promise<Metadata> { const post = await getPost((await params).slug); const image = await ogImageUrl('blog-og', { title: post.title, author: post.author, site: 'yourblog.com' }); return { title: post.title, openGraph: { images: [{ url: image, width: 1200, height: 630, alt: post.title }] }, twitter: { card: 'summary_large_image', images: [image] }, }; } encodeURIComponent leaves ! ' ( ) * unescaped, which RFC 3986 does not — the replace fixes that so the signature matches. We checked this function against the test vector in the API docs. OG_SIGNING_SECRET has no NEXT_PUBLIC_ prefix, so Next.js never inlines it into client bundles; only the signature ends up in the HTML. Two Next.js details to know: - File conventions win. If a route segment has an opengraph-image.tsx or .png, Next.js uses it instead of openGraph.images from generateMetadata — the generateMetadata docs say file-based metadata "has the higher priority and will override" both. Delete the file when you switch. - Absolute URLs. A signed URL is already absolute. If you keep any relative image paths, set metadataBase in the root layout or Next.js cannot build the full URL. Option C: render at build time for static exports With output: 'export' there is no server at request time. A signed URL still works — it is a string computed during the build. If you would rather ship files, call the API from a build script and store the returned URLs: // scripts/og-images.ts — run before `next build` const res = await fetch('https://shareglint.com/v1/images?wait=true', { method: 'POST', headers: { Authorization: `Bearer ${process.env.OG_API_KEY}`, 'Content-Type': 'application/json' }, body: JSON.stringify({ template: 'blog-og', modifications: { title: post.title, site: 'yourblog.com' } }), }); const { url } = await res.json(); // same fields again = same file, no extra render Which to choose | You need | Use | |---|---| | Title, logo, gradient; fonts under 500 KB | next/og | | CSS grid, your own stylesheet, woff2, complex scripts | Real Chrome — signed URL (B) | | No server at all (output: 'export') | Signed URL computed at build, or a build script (C) | | One generic card for every page | A static opengraph-image.png | Whichever you pick, verify the result where it is seen: paste a production URL into LinkedIn's Post Inspector and Facebook's Sharing Debugger — localhost URLs cannot be fetched by either. FAQ What are the limits of next/og ImageResponse? According to the Next.js docs: only flexbox and a subset of CSS (no display: grid), a 500KB bundle limit including JSX, CSS, fonts and images, and only ttf, otf and woff fonts. It renders with Satori and Resvg, not a browser. Can I use Tailwind or my existing React components in an OG image? Only partly with next/og: Satori accepts inline styles and a tw prop, but not your stylesheet, CSS grid or arbitrary components. A renderer that runs real Chrome takes ordinary HTML and CSS, so the card can share your design system. Does opengraph-image.tsx override generateMetadata? Yes. Next.js gives file-based metadata higher priority than config-based metadata, so an opengraph-image file in a route segment replaces the openGraph.images you return from generateMetadata. Do I need the Edge runtime for OG images? No. ImageResponse works on both the Node.js and Edge runtimes, and a signed image URL is just a string you compute — Web Crypto makes the same code work on either. --- Originally published on Shareglint.
Precio del m² en Cajicá por estrato: apartamentos frente a casas (venta, octubre 2026) En Cajicá, el m² de un apartamento en venta va de 4,62 millones de pesos en estrato 2 a 6,72 millones de pesos en estrato 5. En estrato 3, el m² de una casa es 9,5 % más… DOI: 10.5281/zenodo.23521737
MOMENTUM 4 auto-off fix: an app that really switches them off. Testers wanted (5 min) ​ A little while ago I posted here about the MOMENTUM 4 not switching off after the 15 minutes set in the app. On my firmware (3.38.3) it only drops into a standby where Bluetooth stays on, so the battery keeps draining. Wearing them or not makes no difference. The case and a 3 second button hold do switch them off properly, so the firmware can do it. The timer just doesn't use it. What I found: the headphones accept a Bluetooth command that switches them fully off, the same as holding the button. I built ( vibecoded ) a lightweight Android app that sends it when your phone has played nothing and had no call for the time you choose (15 minutes by default). It also does noise control (Adaptive / Custom / Off), EQ with the Smart Control presets, a few other settings, home screen widgets and Quick Settings tiles. It's open source, has no internet permission, and only talks to your headphones over Bluetooth. I'm not affiliated with Sennheiser. Download: github.com/rayane1817/momentum4-auto-off/releases… Code and details: rayane1817/momentum4-auto-off I have only tested it on one pair of headphones (firmware 3.38.3) and two phones (Galaxy Z Flip6 and Nokia XR20). That is why I'm asking for help. What I'd like from you: 1. Install it and use it for a few days. Poke at everything: connect, switch noise modes, drag the EQ faders, try a preset, add a widget or a Quick Settings tile, and turn on the auto power-off guard (Auto power-off tile). Normal use is the best test. 2. If anything doesn't work, looks wrong, doesn't connect, crashes or behaves oddly, copy the logs right then: Device > Copy report, and paste the text in a comment or in the report form (link below). Add one line on what you did and what you expected. Please do it before you close the app, because the log only keeps the most recent lines and starts empty again after a restart. The report has your phone model, Android version, headphone firmware and the last log lines. Bluetooth addresses are removed and a renamed headset's name is hidden. 3. If everything works, that is useful too. One line is enough: phone model, headphone firmware, "works". Questions I'd love answered: - Did "Power off now" switch them fully off (they should not reconnect until you press the button)? - With the guard on, did they switch off after the idle time? - Did they ever switch off while you were wearing them? Especially useful: - Firmware other than 3.38.3 (it's under Device in the app). - Pixel, OnePlus, Xiaomi, Oppo and similar phones that kill background apps aggressively. Does the guard survive? - Other Sennheiser models (MOMENTUM True Wireless 4, ACCENTUM). Please only try "Power off now" and nothing else. The other commands are only mapped for the MOMENTUM 4. - If you'd rather not install anything: your firmware version, your phone, and whether your M4 really switches off after 15 minutes without the app. Things to know: - Once they are off, only the button turns them on again. Bluetooth is off too, so no app can wake them. - If you wear them in silence longer than the idle time, the guard switches them off. Raise the time if that bothers you. - Android will warn you when you install it. It's a self-signed community build. - If it can't connect, force-stop Smart Control Plus first. Only one app can use the control channel at a time. - Use at your own risk. The app refuses the commands that could reset or unpair your headphones, but it's a hobby project. Report form on GitHub (optional, a comment here works too): github.com/rayane1817/momentum4-auto-off/issues/n… And one last question: if you could have one feature next, which would it be?
Precio del m² en Bucaramanga por estrato: apartamentos frente a casas (venta, octubre 2026) En Bucaramanga, el m² de un apartamento en venta va de 4,17 millones de pesos en estrato 3 a 5,48 millones de pesos en estrato 6. En estrato 3, el m² de una casa es 22,4… DOI: 10.5281/zenodo.23521856