AI Daily Brief

Thursday, April 23, 2026

Top stories in artificial intelligence — last 24 hours

OpenAI Ships GPT-5.5 — Faster, Smarter, Pricier

OpenAI released GPT-5.5 today, just six weeks after GPT-5.4 — a turnaround that underscores the accelerating pace of frontier AI competition. The company says the model is its "fastest, sharpest thinker yet" and excels at multi-step workflows including coding, online research, data analysis, and operating software autonomously.

Pricing: $5 / 1M input tokens · $30 / 1M output tokens — double GPT-5.4's rates. Rolls out to Plus, Pro, Business, and Enterprise subscribers in ChatGPT and Codex today.

The model supports a 1M-token context window and will soon be available in the Responses and Chat Completions APIs.

OpenAI Announcement

Google Cloud Next: Gemini Enterprise Agent Platform, 8th-Gen TPUs, A2A at Scale

Google used its Cloud Next 2026 conference to consolidate its AI stack under the Gemini Enterprise Agent Platform — renaming Vertex AI and absorbing Agentspace into a unified product. Key features include a no-code Agent Designer, Google Workspace Studio (agents across Gmail, Docs, Sheets, Drive), and Agentic Data Cloud with cross-cloud Lakehouse capabilities.

8th-Gen TPUs — first ever with two distinct chips engineered specifically for the agentic era. New storage and networking tuned for long context and sequential agent reasoning at scale.

Google's Agent2Agent (A2A) protocol has reached 150 organisations in production — routing real tasks between agents built on different vendor platforms.

Google Cloud Blog

Anthropic Publishes Claude Code Quality Postmortem — Resets Rate Limits for All Users

Anthropic published a detailed engineering postmortem identifying three separate changes responsible for the widely-reported Claude Code quality degradation over the past several weeks:

1. Mar 4 — Default reasoning effort silently lowered from high to medium to reduce UI latency, cutting complex-task intelligence.
2. Mar 26 — Caching optimisation bug cleared model "thinking" on every turn instead of once, causing loss of short-term memory mid-session.
3. Apr 16 — System prompt added verbosity limits (25-word tool responses, 100-word final answers), causing a 3% drop in coding evals.

All three changes have been reverted or fixed in v2.1.116. Anthropic has reset usage limits for all subscribers as of today to compensate for wasted tokens.

Anthropic Engineering

SpaceX Secures $60B Option to Acquire AI Coding Startup Cursor

SpaceX has obtained the right to acquire Cursor — the AI coding editor — for $60 billion later this year. The deal includes a $10 billion "collaboration fee" for joint work regardless of whether the acquisition closes. SpaceX preempted a $2 billion fundraise Cursor was on track to close.

Context: SpaceX's February acquisition of xAI brought Grok, X, and Musk's AI lab under one roof. Pairing Cursor's coding models with SpaceX's Colossus supercomputer would make the combined entity a direct rival to OpenAI Codex and Claude Code in the developer tools market.

The acquisition decision is expected after SpaceX's IPO this summer — the company is avoiding financial-filing updates before listing.

TechCrunch

Claude Code Now Authors 4% of All Public GitHub Commits

New data shows Claude Code wrote roughly 4% of every public commit pushed to GitHub in March 2026 — one in 25 lines of open-source work on the planet authored by a single AI coding agent. A Pragmatic Engineer survey of 906 software engineers put Claude Code at the top of the most-loved AI coding tools with a 46% rating.

The most common developer setup in 2026: Cursor for daily in-IDE editing combined with Claude Code for complex refactors and terminal-based tasks — two complementary tools rather than direct substitutes.

The New Stack

Microsoft Pledges $18B in Australia, Launches First Employee Buyout in 51 Years

Microsoft announced an A$25 billion (~$18B USD) investment in Australian AI infrastructure through 2029, including a commitment to upskill three million Australians in AI by 2028. The announcement came on the same day the company confirmed its first voluntary employee buyout program in 51 years, targeting ~7% of its US workforce.

Scale: Microsoft is targeting $110B–$120B in capex for fiscal 2026 to build Azure AI infrastructure, which grew 39% YoY in Q2. Shares fell ~4% Thursday on the combined news of spending and headcount moves.
Microsoft Source

OpenAI Accuses Anthropic of "Fear-Based Marketing" in Leaked Memo

A leaked memo from OpenAI's Chief Revenue Officer accused Anthropic of overstating its revenue figures to enterprise customers, while CEO Sam Altman publicly accused Anthropic of engaging in "fear-based marketing." The tensions come as Anthropic faces pressure from surging demand straining compute resources, product quality concerns, and pricing disputes.

Axios reported Anthropic's growing pains span multiple fronts simultaneously — compute capacity, security posture, and enterprise pricing — ahead of what analysts describe as a direct showdown for enterprise AI contracts with OpenAI.

Axios

Meta Cuts 8,000 Jobs as AI Infrastructure Spending Hits $115–135B for 2026

Meta announced it will cut approximately 8,000 jobs as part of an effort to offset record AI capital expenditure. The company projected AI-related capex of $115–135 billion in 2026 — nearly double last year's spend — to fund data center buildout and model training at scale.

Meta's Muse Spark model (launched April 8 by its new Superintelligence Labs) is now powering the Meta AI app and rolling out to WhatsApp, Instagram, and Facebook. The workforce reduction reflects a broader Big Tech trend: trading headcount for AI infrastructure.

Bloomberg