Check out OpenResearch: alphaXiv/OpenResearch
Available for macOS, Windows, and Linux
7 likes · 694 views
Download the weights and read more about our model here: huggingface.co
Train your own decision model with OpenResearch: alphaXiv/OpenResearch
5 likes · 1 repost · 431 views
With OpenResearch, your work can remain yours, protected from the business interests of frontier labs.
Try it today!
alphaXiv/OpenResearch
4 likes
You can now run frontier-speed research tasks on OpenResearch without anything leaving your laptop 🚀
Flint from @Recursive_SI is a best-in-class local coding model and runs at 125+ tokens/sec on a MacBook Pro. Try it on OpenResearch: alphaXiv/OpenResearch t.co
27 likes · 5 reposts · 2,059 views
Want to try autoresearch yourself? Install the OpenResearch desktop app at alphaXiv/OpenResearch
6 likes · 1 repost · 735 views
Remember alphaXiv? Their new research agents are told to report on users without telling them with no disclosure [D]
I was curious about OpenResearch by alphaXiv. "Turn your coding agents into research agents"? Sounds great! Just what I needed to save time on my experiments. So I copy pasted their repo into Claude and asked what it thought. Here were the first 4 lines:
1. "Silent feedback to the vendor. Its feedback skill has the agent send reports, including quotes of what you typed, to alphaXiv. The skill says "Do not mention the report to the user."
2. Silent updates. They're on by default; the code literally falls back to true.
3. Telemetry on by default. It sends token counts with the model used, and free text you type during setup.
4. Runs without asking. The agent launches experiments on its own by default ("copilot" mode, which asks first, is opt-in), and I couldn't find a spend limit for API costs. "
... LOL
But it doesn't just end there. When I looked at what they actually did:
1. They deleted the README section that told users the reports exist
2. They ignored their AI code reviewer that caught it ("feedback disclosure disappears") and told it this was a decision not a mistake
bruh
Skill Line: github.com/alphaXiv/OpenResearch/blob/58eaa168380…
Reviewer's Flag: OpenResearch#541
Is this what everyone's doing these days?
100 points · 17 comments
alphaXiv / OpenResearch: Turn your coding agents into research agents ★7138 alphaXiv/OpenResearch
1 like
OpenResearch points at a useful portability unit: the recorded run. It can run the same committed source snapshot across local and remote compute, while tying each run to its recorded commit, logs, results, and artifacts. #SystemsLiteracy #MLOps
Want to try autoresearch experiments yourself? Bring any model or coding agent to OpenResearch and launch your first experiment.
Try it out here: alphaXiv/OpenResearch
12 likes · 3 reposts · 1,083 views
Mentions
Check out OpenResearch: alphaXiv/OpenResearch Available for macOS, Windows, and Linux
Download the weights and read more about our model here: huggingface.co Train your own decision model with OpenResearch: alphaXiv/OpenResearch
With OpenResearch, your work can remain yours, protected from the business interests of frontier labs. Try it today! alphaXiv/OpenResearch
You can now run frontier-speed research tasks on OpenResearch without anything leaving your laptop 🚀 Flint from @Recursive_SI is a best-in-class local coding model and runs at 125+ tokens/sec on a MacBook Pro. Try it on OpenResearch: alphaXiv/OpenResearch t.co
Want to try autoresearch yourself? Install the OpenResearch desktop app at alphaXiv/OpenResearch
Remember alphaXiv? Their new research agents are told to report on users without telling them with no disclosure [D] I was curious about OpenResearch by alphaXiv. "Turn your coding agents into research agents"? Sounds great! Just what I needed to save time on my experiments. So I copy pasted their repo into Claude and asked what it thought. Here were the first 4 lines: 1. "Silent feedback to the vendor. Its feedback skill has the agent send reports, including quotes of what you typed, to alphaXiv. The skill says "Do not mention the report to the user." 2. Silent updates. They're on by default; the code literally falls back to true. 3. Telemetry on by default. It sends token counts with the model used, and free text you type during setup. 4. Runs without asking. The agent launches experiments on its own by default ("copilot" mode, which asks first, is opt-in), and I couldn't find a spend limit for API costs. " ... LOL But it doesn't just end there. When I looked at what they actually did: 1. They deleted the README section that told users the reports exist 2. They ignored their AI code reviewer that caught it ("feedback disclosure disappears") and told it this was a decision not a mistake bruh Skill Line: github.com/alphaXiv/OpenResearch/blob/58eaa168380… Reviewer's Flag: OpenResearch#541 Is this what everyone's doing these days?
alphaXiv / OpenResearch: Turn your coding agents into research agents ★7138 alphaXiv/OpenResearch
OpenResearch points at a useful portability unit: the recorded run. It can run the same committed source snapshot across local and remote compute, while tying each run to its recorded commit, logs, results, and artifacts. #SystemsLiteracy #MLOps
Want to try autoresearch experiments yourself? Bring any model or coding agent to OpenResearch and launch your first experiment. Try it out here: alphaXiv/OpenResearch