alphaXiv's OpenResearch Gains 1,036 GitHub Stars in a Day
A.I. / news
alphaXiv's OpenResearch Gains 1,036 GitHub Stars in a Day
The Stanford-based startup's local-first tool wraps Claude Code, Codex and Cursor, and a two-day-old bug report says its own harness integration doesn't fully work yet.
OpenResearch, an open-source tool from AI research startup alphaXiv, gained 1,036 GitHub stars on Sept. 16, 2026, according to GitHub's trending page, 14 days after the company released it on Sept. 2.
The project, posted on GitHub under the MIT license, had reached roughly 4,200 stars total and 344 commits by Sept. 16. AlphaSignal, an AI-industry newsletter, wrote that OpenResearch turns "coding agents into autonomous scientists" when it covered the release.
What OpenResearch actually automates
OpenResearch wraps existing coding agents, including Claude Code, Codex, OpenCode and Cursor, and gives each one its own isolated git worktree so parallel experiments do not overwrite one another, according to the project's README. An "Autoresearch" mode runs a full propose-code-experiment-inspect loop without a person approving each step, and a git-native experiment tree is meant to keep runs reproducible.
The tool stores projects, conversations and logs in local SQLite files by default, the README says, and can also run over SSH or dispatch jobs to Slurm, Kubernetes, Ray, Hugging Face Jobs, Modal or Tinker. A desktop app is available for macOS and Windows beta, and the CLI installs on macOS and Linux.
| Metric | Figure | As of |
|---|---|---|
| Total GitHub stars | about 4,200 | Sept. 16, 2026 |
| Stars gained that day | 1,036 | Sept. 16, 2026 |
| Open GitHub issues | 8 | Sept. 16, 2026 |
| Days since release | 14 | released Sept. 2, 2026 |
Who built it, and who is paying for it
alphaXiv is better known for a separate product, an interactive discussion layer built on top of arXiv where researchers comment on and rate papers. The company raised a $7 million seed round in November 2025, co-led by Menlo Ventures and Haystack, with Eric Schmidt and Sebastian Thrun among the angel investors, SiliconANGLE reported. Raj Palleti is a co-founder of the company.
OpenAI's own push into agent orchestration takes a hosted approach rather than a local one. It opened its Agents API in beta, putting a similar propose-and-run loop behind a single API call instead of a tool a user runs on their own machine.
The bug that undercuts the pitch
An open issue filed Sept. 14, 2026, says three of OpenResearch's Claude Code harness commands, /compact, /export and /copy, are "getting sent as a message to the chat interface" instead of executing. The same issue asks for /resume support so a Claude Code session can continue inside OpenResearch, which the filer says does not work. As of Sept. 16, the issue had no maintainer reply.
That gap matters because harness integration is the entire premise of the tool: OpenResearch does not run its own model, it drives Claude Code, Codex, OpenCode or Cursor and depends on each one's command surface working as documented. GitHub's own agent-comparison tool found similarly uneven results when it benchmarked wrapped agents against a single strong model directly, beating it on only one of three tests.
What alphaXiv hasn't said
alphaXiv has not published usage numbers beyond GitHub's star and commit counts, and neither the README nor AlphaSignal's coverage states how many people have run a full Autoresearch loop to completion. The company's next visible signal will be whether issue #337 gets a fix or a maintainer response, and whether the star count holds once the launch-week attention passes.
Sources
More in A.I.
- 01GPT-Rosalind-6 Reuses GPT-6 Astra's Weights, Cuts Biology Refusals to 9.4% and Stays Behind a GateOpenAI's Oct. 30 system card addendum says no training was added, so the difference from Astra is deployment controls and who is allowed in.
- 02Architect Says Claude Opus 5.5 Drew a House Set Good Enough for a Permit After Two Redline RoundsStephen Gallagher's informal test of 13 models gave Opus 5.5 the most human feedback, and the drawings still missed a required second exit.
- 0350 GPT Agents in Minecraft Built a Ration Ledger, and 34 Died Doing ItY Combinator's Labs group and Olam Labs ran GPT-6.1 Sol and Claude Opus 5.5 swarms through Minecraft and Mindustry. GPT won both head-to-head games, but each result comes from one harness and one setup.
- 04China's Announcement No. 48 Puts 23 AI Technologies on a Draft Export ListExporters of listed technology must already register with MOFCOM. Comments on the draft catalogue close on November 15, and the sources disagree on whether the thresholds are published.