
GPT-6 Astra Reaches Human-Level Computer Use — and OpenAI Locks the Cyber Door Behind It
OpenAI's GPT-6 Astra hits human-level computer use (72.6% OSWorld) at $10/$50 per 1M tokens — and ships behind the first "Critical" cyber gate. What it means.
Where machine intelligence meets capital
Frontier AI analysis — model benchmarks, infrastructure deep-dives, and the strategy notes that connect new architectures to investable theses.
579 stories · Page 5 of 25

OpenAI's GPT-6 Astra hits human-level computer use (72.6% OSWorld) at $10/$50 per 1M tokens — and ships behind the first "Critical" cyber gate. What it means.

Reports of DeepMind coding research moving to California raise questions about London’s AI cluster. TECHi examines the evidence and Google’s response.

AMD fell 3.7% and Intel 4% while Nvidia held near flat. The market is separating chipmakers that can pass on rising input costs from those that cannot.

Tesla's Cybercab is heading into robotaxi service this month on hardware no customer car has. The fleet that finally drives itself is the one Tesla kept.

ByteDance and Tencent each received about 10,000 Nvidia H200 chips. US licences permit 75,000 per customer. The gap says more than the approval did.

Samsung's foundry share fell to 7.3% from 11.5% in a year, yet it raised advanced chip prices up to 15%, sooner and steeper than TSMC. What that signals.

King’s Cross shows how AI labs turn proximity into economic infrastructure, tightening demand for talent and offices while remote hiring limits the premium.

Anthropic is making Claude Code auto mode the default August 14 after tests showed it beat manual review while still missing some dangerous commands.

OpenAI says Astra may reach its Critical cyber threshold, but key test results remain private. Here is what the disclosure proves—and what it does not.

Firebird opened its Armenia AI factory, but the 70,000-GPU, 300 MW vision still depends on power, financing and customers. Here is what is actually live today.

Kalanick's Atoms raised $1.7B led by a16z, with no valuation disclosed — a bet that physical-AI profit belongs to whoever owns the kitchens, mines and trucks.

Alphabet posted record Q2 2026 net income and its first negative free cash flow together. Why $44.9B of AI capex — not the profit — is the real story.

Black Forest Labs' FLUX 3 puts image, video, audio and robot-action prediction in one model — and holds the open weights back. Here's the market read.

Upper90 will lend General Compute up to $400 million secured by SambaNova inference chips, not Nvidia GPUs. Why inference silicon as loan collateral is an untested bet.

Fireworks AI raised $1.505B at a $17.5B valuation. The number that matters is 95%: nearly all its tokens come from customer-specialized models, not raw open weights.

Tesla fell 14.5% to an 11-month low after Q2 margins hit 1.4%. Post-earnings analyst price targets now split from $130 to $600 — here's what it means for TSLA.

Tesla's Q2 2026 revenue rose 26% to a record $28.24B, but operating income fell 57% and margin sank to 1.4% as robotaxi and AI capex hit $5.79B. What it means for TSLA.

Moonshot AI's Kimi K3, the largest open-weight model ever, opens July 27. It's cheap per token but needs ~1.4TB resident — here's who can really run it.

OpenAI, Meta, NVIDIA/ServiceNow, and Google shipped enterprise agent platforms in a month — and disagree on what the product is. The market-structure read for buyers.

Google shipped Gemini 3.6 Flash, 3.5 Flash-Lite, and a restricted 3.5 Flash Cyber model. Why the gated cybersecurity variant is the real signal in the release.

Palantir stock fell about 6% on July 22 as NHS scrutiny, an open-source tool, and valuation nerves converged. Why the real test is the August 3 earnings print.

EU AI Act enforcement is live. See who is exposed, how complaints work, what Article 50 requires, and why GPAI fines can reach 3% of revenue.

LangChain now standardizes reasoning effort across OpenAI, xAI and Fireworks, but identical labels still hide different cost, latency and model behavior.

Claude Code 2.1.216 targets permission drift across resumed agents, isolated worktrees, rewind, and sandbox controls. Here is what teams should test.