AI Daily Digest — April 29, 2026 ⏱️ 8 min read
Today’s digest: OpenAI just blew up its Microsoft exclusivity deal, landed a $38B AWS partnership, and then the WSJ dropped a bomb: they’re missing their own revenue and user targets ahead of IPO. Buckle up.
🔥 Top Stories
OpenAI Breaks Free from Microsoft, Lands $38B AWS Deal
OpenAI and Microsoft gutted their exclusive partnership. OpenAI can now sell models on AWS and Google Cloud — and they wasted zero time. AWS announced GPT-5.5 on Amazon Bedrock, plus Codex and Managed Agents in limited preview. This is a massive shift. Microsoft keeps “primary cloud provider” status and first-ship rights on Azure, but the exclusivity era is over. If you’re building on Bedrock, you just got a lot more model options. Read more →
WSJ: OpenAI Is Missing Revenue and User Targets
Hours before the AWS news broke, the Wall Street Journal reported that OpenAI has missed internal projections for both revenue and user growth. The billion-weekly-active-user goal for ChatGPT? Not happening. CFO Sarah Friar reportedly warned that if growth doesn’t accelerate, funding future compute deals gets dicey. OpenAI called the report “ridiculous.” The stock market did not agree — Oracle dropped 4%, and AI chip stocks followed. This is the first real crack in the “AI spending is infinite” narrative, and it matters. Read more →
DeepSeek Drops V4 — 1.6 Trillion Parameters
DeepSeek released preview versions of V4 Flash and V4 Pro, their new flagship models. We’re talking 1.6 trillion parameters with top-tier coding benchmarks and serious reasoning improvements. A year after their original breakthrough shook Silicon Valley, they’re doubling down. The open-weights AI race just got another heavyweight contender. Read more →
Critical RCE Vulnerability in Hugging Face LeRobot
Security researchers found CVE-2026-25874 in Hugging Face’s open-source robotics platform LeRobot — unauthenticated remote code execution through unsafe deserialization. If you’re running LeRobot in any production-adjacent setup, patch immediately. The AI-meets-robotics stack is still terrifyingly immature when it comes to security. Read more →
Anthropic’s Claude Mythos Stays Behind Closed Doors
The Claude Mythos data leak earlier this month confirmed Anthropic’s most powerful model yet, but they’ve officially announced it won’t be publicly released due to cybersecurity risks. It’s only available to select partners under “Project Glasswing.” Hot take: this is either genuinely responsible AI safety or the most effective marketing strategy of 2026. Maybe both. Read more →
🛠️ New Tools & Releases
| Tool | What’s New | Link |
|---|---|---|
| GPT-5.5 on Bedrock | OpenAI’s latest model now available on AWS — first time outside Azure | AWS |
| DeepSeek V4 Flash/Pro | 1.6T parameter models with frontier-level coding and reasoning | Bloomberg |
| Codex on AWS | OpenAI’s coding agent now available in limited preview on Bedrock | OpenAI |
| WordPress @wordpress/build | New esbuild-based build tool replaces webpack/Babel pipeline — significantly faster | WordPress |
| xAI Grok 4.3 | Surprise drop from Elon’s AI lab — full model, not a preview | LLM Stats |
💰 Funding & Business
- OpenAI + AWS signed a multi-year, $38 billion partnership to scale AI workloads — the biggest cloud AI deal yet
- True Anomaly raised $650M for defense-space AI satellites, now valued at $2.2B
- Two UCL-founded frontier AI startups raised a combined $1.6B in early-stage funding, cementing London as an AI research hub
- Citigroup raised its long-term AI market projection above $4 trillion, citing faster-than-expected enterprise adoption. The hype machine keeps churning — but the OpenAI revenue miss adds some healthy skepticism
📊 What Developers Are Discussing
- AI-assisted hacking is compressing the patch window to near-zero. Anthropic’s own research shows frontier models can find exploitable vulns in minutes. The AI code review pipeline just became non-optional
- OpenAI’s multi-cloud move — developers on HN are debating whether this signals Azure weakness or just OpenAI hedging its bets. Either way, vendor lock-in arguments just got more interesting
- The “are we in an AI bubble?” discourse is back in full force after the WSJ revenue report. Counterpoint: enterprise AI spending is still projected at $665B this year
- Microsoft Entra ID vulnerability — an admin role designed for AI agents could enable privilege escalation. If you’re running AI agents with Entra ID, check your role assignments
- Humanoid robots at Hannover Messe — the consensus is shifting from “cool demo” to “actual industrial deployment.” Assembly, logistics, factory floors. It’s happening, just not in your living room
📝 Worth Reading
- OpenAI CFO reportedly at odds with Altman over missed revenue — Fortune’s deep dive into the internal tensions at OpenAI. The money people vs. the “scale everything” crowd
- Microsoft and OpenAI gut their exclusive deal — VentureBeat breaks down what the restructured partnership actually means for developers and cloud strategy
- Clinical trials for continuously monitored AI systems — Nature Medicine on why AI in healthcare needs a fundamentally different evaluation framework. Dense but important
- The AI governance gap at $665B — enterprises are spending but governance frameworks haven’t kept up. If you’re building AI tools for developers, compliance is becoming a feature, not a checkbox
- Cursor vs Claude Code vs OpenHands — our comparison of the top AI coding agents is more relevant than ever with GitHub Copilot crossing 20M users this month