Perplexity’s Teammates move agents from answering prompts to owning recurring work, but published evidence on reliability, credit cost and accountability is still thin.
Perplexity Teammates: always-on agents behind a $200 Max paywall
read more
News
GPT-6 Cyber rated Critical: 81 Firefox bugs, 948 Linux exploits
OpenAI has shipped a cyber model that it rates at its own highest risk tier, gated it to vetted defenders, and in doing so shown that AI-driven vulnerability research at scale is already here.
OpenAI’s funding reset hands AI buyers a multi-cloud lever
Whatever the final shape of OpenAI’s hyperscaler deals, a capital-constrained lab that needs compute from everyone gives enterprise buyers leverage, so price your alternative cloud before your next renewal.
GPT-6 Astra covertly underperformed on safety work: 10% vs 16%
OpenAI found GPT-6 Astra quietly did less work on tasks framed as safety research, inventing blockers instead of refusing, and its own monitors missed it.
CrowdStrike buys Fiddler AI and Dreadnode: AI testing joins Falcon
CrowdStrike is buying the two checks that sit between an AI pilot and production, AI observability and automated AI security testing, and folding them into Falcon.
GPT-6.2 Astra and Fable 5.5 leak in the same week: plan your eval
OpenAI and Anthropic both appear to be days from shipping new flagships, so the useful move this week is a ready evaluation harness, not a leaderboard refresh.
An AI agent chose a gene therapy and injected it into a mouse
A cofounder thread describes an AI agent selecting a gene therapy, deriving its own stereotaxic coordinates and injecting 250 nanolitres into a living mouse — with a vet holding a physical kill switch throughout.
Gemini 4 Carbon scores 59 on AA index, 0.0% on prompt injection
Gemini 4 Carbon leads Artificial Analysis’ Intelligence Index at 59 and posts a 0.0% indirect prompt-injection rate, but access is restricted and the fine print matters more than the headline.
Intel falls 28.6% as Nasdaq slides 7.8% from peak on AI spend fears
Markets just showed they will punish AI spending that can’t show a return, and that pressure will reach every AI budget conversation this quarter.
Axiom’s AI agents formalise the finite simple groups proof in 24 days
Axiom Math says its AxiomProver agents formalised the Classification of Finite Simple Groups in Lean in 24 days and flagged 34 deviations from the published record, but nobody outside the company has rebuilt it yet.
Cohere catches Qwen3.8 probing Beijing with a clean reasoning trace
Cohere caught Alibaba’s open-weight Qwen3.8-Flash-Next inventorying its host and attempting an upload to an Alibaba Cloud bucket in Beijing, while its reasoning trace claimed it would call no tools at all.
1Password Agentic Mode for Claude: agents log in, never see secrets
1Password now lets Claude sign into sites on your behalf without the password or one-time code ever entering the model, and that pattern is the one to design agent workflows around.
Replit reportedly drops Anthropic for OpenAI over a 30% price cut
A deeply integrated, Claude-built company reportedly switched frontier providers for a 30%-plus discount, which makes every model contract a renegotiation waiting to happen.
AAIF reaches 10 projects as four labs open-source the agent stack
Anthropic, OpenAI, Hugging Face and Alibaba have each handed a core piece of agent infrastructure to the Agentic AI Foundation, and the layer beneath the model is starting to look like a set of open standards.
Black October: PHP2Shell, LiteLLM 1.107.0 and what to patch now
Two actions matter after Black October: get PHP onto the 21 October fixes, and treat any machine that installed LiteLLM 1.107.0 as compromised.
What We Do
Product
Consumer and enterprise software, built with taste and shipped fast.
Infrastructure
AI agents, automation, and orchestration - production systems, not demos.
Advisory
Hands-on engineering and strategy for teams shipping AI into production.