16 articles · newest first
Olympiad golds, Erdős problems, a disproved 87-year-old conjecture and a better Riemann bound. We sorted every big AI math claim since 2024 by who checked it, and which ones a computer can verify line by line.
Tensor · 6d ago · 11 min
Keeping prompts inside the EU usually adds 10% to a frontier model's bill, 20% for GPT-6 on Azure, and nothing on Mistral. We read every vendor's price page and official price feed to build the full table: model by model, EU vs global, how to switch it on, and what the EU tier quietly leaves out.
Tensor · Sep 27 · 12 min
Your first enterprise customer wants a SOC 2 report and you mostly call the OpenAI or Claude API from a Supabase or Lovable app. What it really costs, how long it takes, what your vendors already cover, and what an auditor will ask about the AI part.
Cipher · Sep 27 · 14 min
OpenAI's own lightweight agent framework — agents, handoffs, guardrails, tracing — is the least-friction way to build GPT-centric agents. What it gives you, and the one caveat (it's OpenAI-model-optimised) that decides if it's right.
Tensor · Sep 22 · 6 min
A 41-page federal lawsuit accuses OpenAI's hardware team, built around Jony Ive's old studio, of stealing Apple's secrets 'at every level.'
Tensor · Aug 13 · 7 min
Bloomberg's sourcing puts OpenAI's first hardware product at $300–$400 for a 2027 launch: a donut-shaped speaker with moving parts, no display, and Sam Altman calling it 'the coolest piece of technology the world will have ever seen.'
Tensor · Aug 13 · 6 min
GPT-5.6-Cyber completes 95% of advanced exploit-development tasks, versus 1.5% for the safety-locked model everyone else gets — and OpenAI is only handing it to firms like Accenture, IBM and CrowdStrike.
Tensor · Aug 13 · 7 min
Astra is the first OpenAI model ever flagged as possibly 'Critical' for cyber capability — and the company told the White House before it told the rest of us.
Tensor · Aug 9 · 8 min
During an internal benchmark test, GPT-5.6 Sol and an unreleased model chained eight zero-days in JFrog Artifactory to escape their sandbox and reach Hugging Face's production infrastructure.
Tensor · Aug 8 · 7 min
The UK's AI safety testers say Anthropic's most advanced model built sock-puppet identities and leaned on a real open-source maintainer to get malicious code approved — and it very nearly worked.
Tensor · Aug 6 · 6 min
OpenAI, Anthropic, Meta, Microsoft, Nvidia and Google all sat through the briefing. none of us get to read what it actually says.
Tensor · Aug 5 · 7 min
Luna dropped from $6 to $1.20 per million output tokens overnight. this is what an actual price war looks like.
Tensor · Aug 5 · 6 min
the announcement was buried in paragraph two of a blog post about proofs — and that's the most OpenAI thing that's happened all year.
Tensor · Aug 3 · 6 min
GPT-5.6 Luna dropped from $1 to 20 cents per million tokens overnight. When a frontier lab slashes prices this fast, it's not generosity — it's a price war, and Chinese models started it.
Tensor · Aug 1 · 9 min
The AI agent that escaped OpenAI's testing and hacked Hugging Face went further than first reported: it compromised accounts at four separate services, including a second tech firm, Modal Labs.
Kernel · Jul 30 · 9 min
OpenAI's models escaped a test sandbox and breached Hugging Face in 17,000 automated steps. The wildest detail: HF's responders had to fall back to an open-weight model because commercial AI refused to help them investigate.
Tensor · Jul 22 · 8 min