Filters
- OpenAI scraps October release of GPT-6.1 Astra after internal safety tests, OpenAI confirms
OpenAI has scrapped GPT-6.1 Astra, which was planned for an October debut in ChatGPT and Codex, after internal testing found it did not meet the company's safety and alignment standards, Reuters reports. The Wall Street Journal first reported the decision, and OpenAI confirmed it. Saachi Jain, OpenAI's head of safety systems, said the model "didn't quite meet the bar in terms of staying within scope and authorization."
- Corral, a Linux CLI, makes sure no process started by an AI agent's command is left running
Corral runs a command with a time limit and, before it returns, verifies that every process the command started has exited, returning exit code 120 if it cannot prove that. Its developer built it for AI coding agents and CI jobs that leave behind background jobs, daemons or processes holding output open; it uses cgroup v2's cgroup.kill when it can and otherwise acts as a child subreaper. Version 0.1.0 needs Linux 5.11 or later and is not a security sandbox.
- Cline was featured at Meta Connect!
Cline was featured at Meta Connect! We are one of the best harnesses to use their Muse Spark 1.3 model, and their contributor variant is completely free through Cline provider. Incredible seeing Meta make a comeback 🔥 https://t.co/glcRpMcvd7
- Anthropic releases Claude Sonnet 5.5, 30%+ faster than Sonnet 5 at the same price
Anthropic says Claude Sonnet 5.5, the second model in the Claude 5.5 family, generates output more than 30% faster than Sonnet 5 and costs up to 30% less per task for most work, at the same $2 per million input tokens and $10 per million output tokens. Anthropic says it scores 70.6% on Terminal-Bench 4.0, versus 10.3% for Sonnet 5, and lands two points below Opus 5.5 on GDPval-AA. It is available now, including on AWS, Google Cloud and Microsoft Azure and on the Claude Platform as claude-sonnet-5-5, and Claude Haiku 5.5 is due in the coming weeks.
- Claude Code 2.1.284 makes Sonnet 5.5 the default model
Claude Code v2.1.284 adds Claude Sonnet 5.5 (claude-sonnet-5-5) as the default Sonnet model on the Anthropic API, per the GitHub release notes. The release notes list 1M context for the model at $2/$10 per million tokens with $0.20 per million cache reads.
- Claude Sonnet 5.5 one-shots a 3D zombie FPS in 49 minutes for $1
A Reddit post shows Claude Sonnet 5.5 one-shot an entire 3D zombie FPS game in 49 minutes at a cost of $1.
- Claude Opus 5.5 is now available
Claude Opus 5.5 is now available, per Anthropic's site. AWS's Kiro team describes it as Anthropic's most capable Opus model yet, delivering Fable 5.1-level performance on most work at 40% lower cost.
- ChatGPT Health now explains a chart, metric or record you select, based on your own data
OpenAI's ChatGPT release notes say users can now select a chart, metric or record in the Health tab to get personalized explanations and insights based on their connected health information. Health is available to logged-in Free, Go, Plus and Pro users in the U.S. aged 18 or older on web and iOS, and is rolling out gradually.
- AutoTrust AI releases JEV-27B, an open-weights decision model for self-hosted AI agents
AutoTrust AI's JEV-27B, released under Apache-2.0, adds a 108.9M-parameter decision block to a frozen Qwen3.8-27B so it can answer yes/no, multiple-choice and 0-5 rating questions in one forward pass with a calibrated probability for each option, while leaving the base model's generation path unchanged. The company says the block trained in about 9.2 hours on one NVIDIA B200, and that JEV-27B averaged 84.07% on six public decision benchmarks versus 83.85% for TypeSafe's hosted Jev 1.13, the model it was distilled from; AutoTrust ran both comparisons itself.
- Anthropic, Gamma, and Clay on taking enterprise AI beyond the demo, per TechCrunch
TechCrunch reports the three companies shared at TechCrunch Disrupt 2026 what it takes for an AI product to go beyond the demo in real enterprise deployment.
- AMD to acquire Fei-Fei Li's World Labs in an all-stock deal valued at about $8.2 billion
AMD says it has signed a definitive agreement to acquire World Labs, the San Francisco lab led by Fei-Fei Li that builds spatial-intelligence models that generate, reconstruct and simulate interactive 3D environments from text, images and video. The all-stock deal is valued at about $8.2 billion and is expected to close by the end of 2026, subject to regulatory approvals. After closing, Li will join AMD as executive vice president and chief scientist, reporting to CEO Lisa Su.
- Engram, a sampler that runs small AI audio models offline, launches on Kickstarter, per The Verge
The Verge reports that music startup Thoughtful Things has opened a Kickstarter for Engram, a sampler and groovebox that uses small local AI models to transform input audio and generate new sounds without an internet connection. The company says its custom models are trained only on openly licensed, commercially usable datasets such as CC-BY. Pledges start at $675 on Kickstarter; shipping dates were not reported.
- Thales expands its Google Cloud partnership, integrating AI Security Fabric with Gemini Enterprise to secure agent workflows
Thales says its AI Security Fabric now integrates with Google Cloud's Gemini Enterprise to provide real-time security, governance and visibility over interactions between users, AI agents, models and tools. It says the setup can set boundaries for agents, limit their data access and block unsafe actions as they happen, targeting risks such as prompt injection, sensitive-data leaks and unauthorized actions.
OfficialSource: Business Wire - TextCLF launches TQ, calibration-free 4-bit quantization with an OpenAI-compatible API
TextCLF says its Quant Factory converts pretrained models to TQ 4-bit models without a calibration dataset; the surrounding pipeline is open source, while the quantization and inference core is closed. Models run via an OpenAI-compatible API or self-hosted with Docker and vLLM, and TextCLF says Qwen3.8 27B shrank 4.1x while keeping 99.6% of FP16 accuracy, a figure not independently verified.
OfficialSource: TextCLF on X - Researcher says OpenAI agents hit a UN statistics site 16,000+ times, per The Verge
The Verge reports that security researcher Rowan Howard-Jones says OpenAI agents scanned UNCTAD's statistics website more than 16,000 times between April and June, apparently seeking public Productive Capacities Index data without API access, and later worked around HTTP tool restrictions to pull it. The report includes no response from OpenAI.
- OpenAI fixes a bug that degraded image understanding in GPT-6 Sol and GPT-6 Luna
OpenAI Developers says the fix should improve vision tasks, including computer use, in the API and Codex.
- OpenAI shuts down gpt-3.5-turbo-instruct, babbage-002 and davinci-002 today
OpenAI's deprecations page lists September 28 as the shutdown date for gpt-3.5-turbo-instruct, gpt-3.5-turbo-1106, babbage-002 and davinci-002, and recommends moving to gpt-5.6-terra. The shutdown is part of the legacy GPT snapshot retirement OpenAI announced in September 2025.
OfficialSource: OpenAI deprecations - Codex CLI 0.158.0 adds MCP OAuth client secrets and transparent image output
Codex 0.158.0 (stable) can connect to MCP servers that require a pre-registered OAuth client secret, and can generate images with a transparent background. It also protects the exec-server WebSocket with a bearer token and adds select-to-copy and right-click paste in the full-screen TUI. It fixes several Windows and Linux sandbox issues.
OfficialSource: OpenAI Codex release - Users report Codex usage limits draining unusually fast
A ChatGPT Plus user on the OpenAI Developer Community forum says Codex now exhausts its usage window far faster than before, with Sol Medium running out after about nine minutes of visible work. Other forum posts report similar issues since September 26; OpenAI has not explained the cause.
- NVIDIA unveils Open Agent Safety Platform, pairing open-source OpenShell with Sentry on BlueField-4
NVIDIA's developer blog describes a reference design that runs AI agents in the open-source OpenShell sandbox (Apache 2.0) under verifiable policies, paired with NVIDIA Sentry on BlueField-4 DPUs for out-of-band, in-silicon monitoring and policy enforcement. NVIDIA says it built the platform with more than 100 industry partners; OpenShell itself was released earlier.
OfficialSource: NVIDIA Developer blog