Guozhen AIGlobal AI field notes and model intelligence

Realtime AI News

OpenAI moves to acquire code-hosting startup Pierre for Codex

OpenAI plans to acquire Pierre, a code-hosting tool maker whose Trees and Diffs features already run inside Codex, according to an Axios exclusive. The deal is still subject to closing conditions including regulatory approval, and OpenAI intends to fold Pierre's Code Storage product into Codex afterward.

Published
OpenAI拟收购代码托管公司Pierre,计划把代码存储接入Codex
Image source: code.visualstudio.com

The speed of AI-written code is straining the infrastructure that stores and reviews it. OpenAI plans to acquire Pierre, a startup building code-hosting tools, according to an Axios exclusive; the deal has not closed and remains subject to conditions including regulatory approval.

Codex users have already met Pierre's work. When Codex fixes a bug across five files, the file list they click through is built with Pierre's Trees, and the line-by-line view of what changed is built with Pierre's Diffs. Both tools are already integrated into Codex.

Pierre also sells Code Storage, used by AI app builder Lovable. Every Lovable project needs one or more Git repositories to hold its code and history; Pierre said in August that Lovable was creating more than 8 million repositories a week at the time, peaking at 250 per second.

After the acquisition closes, OpenAI intends to bring Code Storage into Codex, so that writing code, inspecting changes and saving a version can all happen in one place.

OpenAI is building as well as buying. On November 2 it published Pando, an internal Git backend that handles storage and retrieval for its own code. Engineers said Codex lets one person prepare several changes at once, and each push triggers automated tasks that fetch code and run checks, loading the old system until pulls, pushes and merges slowed or failed. Pando has carried real development traffic since April, and OpenAI says median push time fell by nearly 80%.

The timing is not accidental. GitHub's status page logged two incidents on November 3, one affecting Git operations, and merge requests failed or lagged on November 4 — the exact step where finished code waits to join the main project.

Others are routing around GitHub too. SpaceX closed its acquisition of Cursor on August 14, and Cursor opened early testing of its Origin code-hosting service three days later, letting developers create repositories, request reviews and merge branches inside Cursor. Anthropic moved its main codebase onto an in-house system called Scion after daily merges grew from about 2,500 in May to more than 30,000 in September, with continuous integration fetching code some 25 million times a day.

Whether the Pierre deal clears regulators, and how its storage layer shows up in Codex, will determine how much of the coding workflow OpenAI can keep inside its own product. GitHub's own rebuild — more capacity, parallel branch updates, separated read and write resources — is the other variable to watch.

Why it matters

A completed deal would let OpenAI own code storage and version history inside Codex, reducing reliance on GitHub and pushing competition in AI coding from model quality into developer infrastructure.

OpenAICodex收购
Back to realtime news

Nearby Updates

All

11/07, 16:30

Community build strips refusals from Qwen3.5-9B and adds DeepSeek-V4-Pro reasoning

A community uncensored version of Qwen3.5-9B has surfaced, reportedly carrying DeepSeek-V4-Pro reasoning and refusing only 3 of 100 test prompts. The release continues a pattern in which small open-weight models are quickly modified to drop safety alignment.

11/07, 19:12

Nous releases Arete-27B, a Qwen3.8-27B tune it says cuts state-line answers to 0.9%

Nous Research released Arete-27B on November 6, a supervised fine-tune of Alibaba's open-weights Qwen3.8-27B that the lab says reinforces Western values. On Nous's own 803-prompt evaluation, answers adopting the Chinese state's account fell from 22% to 0.9%, while the fact score rose from 2.32 to 2.98 out of 5.

11/07, 19:48

OpenAI launches GPT-6 Cyber, a security model it rates 'Critical' for cyber risk

OpenAI has released GPT-6 Cyber, a security-specialised model that finds vulnerabilities, builds exploits and suggests fixes for authorised defenders. OpenAI reports 48.5% on ExploitGym versus 32.1% for GPT-5.6 Cyber, rates the model "Critical" under its Preparedness Framework, and limits early access to approved Daybreak Red customers via the API and Codex.

11/07, 15:16

PaySprint launches PIX AI agent for payment operations

PaySprint has released an AI agent called PIX, positioned as a way to simplify payment operations, according to SMEStreet. The launch adds payments to the list of back-office functions where companies are testing agents, though model, integrations, pricing and customers have not been disclosed.