Everything that mattered in AI, one page a week.

OpenAI announced an agreement to acquire Astral, the Rust-based Python tooling company behind uv, Ruff and ty, with the team joining OpenAI's Codex group.

Astral's Charlie Marsh wrote that the tools have grown to hundreds of millions of monthly downloads and that OpenAI will continue supporting the open-source projects after close. The deal is subject to regulatory approval; until it closes, the companies remain independent. All three tools are permissively licensed, so existing releases cannot be retroactively relicensed.

MiniMax released M2.7, a sparse MoE agentic coding and reasoning model of about 230B total and 10B active parameters with a 205k context, whose headline claim is early echoes of self-evolution.

An internal version of M2.7 acted as an autonomous researcher inside MiniMax's own reinforcement-learning pipeline, reading logs and loss curves, debugging data-pipeline failures, adjusting hyperparameters and proposing its next experiments through a three-part harness: markdown short-term memory, self-critique and self-optimization. MiniMax says that covered an estimated 30–50% of the RL workflow, with humans making the key decisions, and reports 56.22% on SWE-Pro and a 66.6% medal rate on MLE-Bench Lite after a 25-hour recursive-evolution run. M2.7 also ships native Agent Teams support.

OpenAI released GPT-5.4 mini and nano, described as its most capable small models, targeting high-volume workloads and subagents.

mini runs more than 2x faster than GPT-5 mini and approaches the full GPT-5.4 on some evals — 72.1% on OSWorld-Verified versus 75.0% for GPT-5.4 and 42.0% for GPT-5 mini. nano is API-only at $0.20 per 1M input and $1.25 per 1M output tokens, undercutting Gemini 3.1 Flash-Lite on input. In ChatGPT, mini is available to Free and Go users and serves as the rate-limit fallback for GPT-5.4 Thinking on paid tiers.

At GTC 2026 in San Jose, Jensen Huang said purchase orders for Blackwell and the next-gen Vera Rubin platform will reach $1 trillion through 2027 — double the $500B opportunity Nvidia projected a year earlier.

Vera Rubin was detailed as a seven-chip, five-rack-scale-system platform (Vera CPU, Rubin GPU, BlueField-4 STX, NVLink 6, ConnectX-9, Spectrum-6, Groq 3 LPU), with Nvidia claiming up to 10x lower cost per token and 3.3x the inference performance of Blackwell Ultra GB300 NVL72 — vendor figures, not independent benchmarks.

The software half matters as much: NemoClaw, an open-source stack for OpenClaw agents (Nemotron open models plus the OpenShell sandbox runtime), and the Nemotron Coalition of six frontier open model families spanning language, world and vision, robotics, driving, biology and weather.

Mistral released Small 4 under Apache 2.0, its first model to merge Magistral, Pixtral and Devstral into a single 119B-parameter MoE with 6B active parameters per token and a 256k context window.

A new reasoning_effort parameter toggles between none for low-latency chat and high for step-by-step reasoning. Mistral reports a 40% reduction in end-to-end completion time in its latency-optimized setup and 3x more requests per second versus Mistral Small 3, at $0.15 per 1M input and $0.60 per 1M output tokens. It shipped day-0 as an NVIDIA NIM.

A free, unattributed model called Hunter Alpha appeared on OpenRouter, which the platform labeled a stealth model. Its profile advertised roughly 1 trillion parameters, a 1M-token context window, reasoning capability and free access — a combination Reuters noted matched widely reported expectations for DeepSeek's unreleased V4.

It rapidly topped OpenRouter's daily usage charts, passing 500 billion tokens in its first week. At the time nobody had claimed it; Xiaomi's MiMo team later confirmed it was an early internal test build of MiMo-V2-Pro, its flagship agent model.

Google LLC completed its $32 billion all-cash acquisition of cloud and AI security platform Wiz, a year after announcing it and following US DOJ clearance in November 2025 and unconditional EU approval on February 10, 2026.

Wiz joins Google Cloud but keeps its brand and its commitment to securing AWS, Azure and Oracle Cloud environments. Wiz had crossed $1 billion in ARR in 2025, roughly double the run rate at the time of Google's first, rejected $23 billion approach. The close makes it both the largest acquisition in Google's history and the largest-ever acquisition of a venture-backed startup.

NVIDIA and Mira Murati's Thinking Machines Lab announced a multi-year strategic partnership to deploy at least one gigawatt of next-generation Vera Rubin systems for frontier model training, with deployment targeted for early next year.

NVIDIA also made a significant undisclosed investment in the startup, and the two will co-design training and serving systems for NVIDIA architectures. At NVIDIA's own stated ~$50B-per-gigawatt figure, the compute commitment alone is worth tens of billions.

Anthropic filed two lawsuits against the Pentagon — one in the Northern District of California and one at the DC Circuit Court of Appeals — challenging the Department of War's designation of the company as a supply chain risk to national security.

The fight stems from Anthropic's refusal to drop two contractual red lines: no domestic mass surveillance and no fully autonomous lethal targeting. The designation, formalized in a March 4 letter, was the first time the tool had been aimed at a US company, and it threatened to force any government contractor doing business with the Pentagon to cut ties with Anthropic. Anthropic argues the move violates the First Amendment and is an arbitrary abuse of procurement authority; government filings later claimed concern that Anthropic might preemptively alter Claude's behavior mid-warfighting.

Advanced Machine Intelligence Labs, the Paris-based startup co-founded by Turing Award winner Yann LeCun after he left Meta in November, closed a $1.03 billion seed round at a $3.5 billion pre-money valuation.

The round was co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions, with Nvidia, Mark Cuban, Eric Schmidt and Tim Berners-Lee among the backers. AMI is building world models on LeCun's JEPA approach — systems that learn from real-world sensor data rather than text — and is hiring across Paris, New York, Montreal and Singapore. LeCun has said the lab will be one of the few frontier labs that is neither Chinese nor American.