Morning Digest, May 1, 2026
May 1, 2026 · 17 newsletters · 5 overlapping stories
Top Stories
Big Tech Posts $112B Earnings Quarter — AI Capex Hits Record
(5 newsletters)
Alphabet, Amazon, Meta, and Microsoft all reported blowout Q1 results, collectively spending $112 billion in capital expenditures as the AI land grab accelerates. Google Cloud grew 63% to $20B with a $460B backlog. AWS grew 28%, its fastest in 15 quarters, and processed more Bedrock tokens in Q1 than in all prior years combined. Meta delivered 33% revenue growth while raising capex guidance as high as $145B for 2026. Microsoft’s Copilot crossed 20 million paid seats and its AI annualized revenue hit $37B, up 123% year over year. Markets cheered Alphabet (+10%) while Meta fell about 6% and Microsoft about 2% on investor unease over spending pace. The Nasdaq closed its best month since April 2020.
Elon Musk’s Trial Testimony Puts OpenAI’s Future in Play
(5 newsletters)
Day two of Musk vs. Altman produced fireworks: Musk testified under oath that xAI used distillation techniques to train Grok on OpenAI models, and called himself a “fool” for donating $38 million to a startup now worth $800 billion. The four-week trial could force Altman off the board and disrupt OpenAI’s looming IPO. Musk is seeking over $180 billion in damages and wants the court to unwind the company’s for-profit conversion. Separately, Altman made a notable claim to The Atlantic about synthetic data that observers say could reshape how the next generation of AI gets trained.
ChatGPT’s Goblin Obsession: The Origin Story
(4 newsletters)
OpenAI traced ChatGPT’s habit of inserting goblins, gremlins, and fantasy creatures into unrelated conversations back to a single reward signal in its “Nerdy” personality preset. After ChatGPT-5.1 launched last November, goblin mentions jumped 175% in user conversations. The Nerdy preset drove two-thirds of all goblin mentions from just 2.5% of traffic, and fine-tuning loops recycled the creature-favored outputs into the default model. OpenAI retired the preset in March and shipped GPT-5.5 with a Codex system prompt explicitly banning goblins, gremlins, ogres, trolls, raccoons, and pigeons.
Zuckerberg’s Biohub Commits $500M to AI Biology
(4 newsletters)
Mark Zuckerberg and Priscilla Chan’s Biohub nonprofit announced a $500M Virtual Biology Initiative aimed at building open datasets large enough to let AI simulate how disease starts at the cellular level. Of that, $400M funds Biohub’s own data generation and imaging work, with $100M going to external research labs. Nvidia, the Allen Institute, and Arc are joining the effort. Biohub’s lead researcher noted current AI biology datasets max out near 1 billion cells, and an “order of magnitude” more is needed. The goal: train models to understand disease and reprogram it “at the level of cells, molecules, and tissues.”
Cursor Releases TypeScript SDK — AI Agents as Developer Infrastructure
(4 newsletters)
Cursor released a public beta TypeScript SDK that exposes the same agents, runtime, and models powering the Cursor desktop app to any developer workflow. Teams are already using it in CI/CD pipelines to auto-fix build failures and file PRs. Developers can swap between Claude, GPT, and other frontier models with a single line of code and hook into MCP servers. A related wrinkle: Anthropic’s Opus 4.7 tokenizer breaks prompts into more pieces, quietly raising costs 12 to 27% for longer prompts, and Claude’s weekly usage caps are pushing some developers back to OpenAI Codex.
Also Worth Knowing
-
The White House Complicates the Anthropic-Pentagon Standoff. Anthropic wanted to expand access to Mythos from 50 firms to 120. The White House pushed back over compute concerns while a new national security memo prepares to address some of Anthropic’s grievances. Secretary of War Pete Hegseth still called Anthropic “run by an ideological lunatic,” so internal alignment is fragile.
-
Netflix Launches Clips, Its TikTok-Style Vertical Feed. Rolling out to the US and eight other countries, Clips serves short clips from Netflix’s content library to help users discover shows. Plans include podcasts, live programming, and genre-based collections. Peacock and Disney+ recently launched similar features.
-
Gemini Can Now Generate Files Directly in Chat. Google’s Gemini app gained the ability to output fully formatted Docs, Sheets, Slides, Word, Excel, CSV, Markdown, LaTeX, and PDF files from a single chat prompt, globally for all users.
-
Apple’s iOS 27 Will Overhaul AI Photo Editing. Three new tools coming to iPhone: Extend (generative background expansion), Enhance (auto lighting and quality), and Reframe (perspective shifting on spatial photos). Expected reveal at WWDC on June 8.
-
AWS Opens Its Doors to OpenAI Models via Bedrock. OpenAI’s latest models, Codex, and production-ready agent tooling are now available inside Amazon Bedrock with AWS-native security and governance. The deal effectively ends any remaining notion of Microsoft exclusivity and doubles infrastructure options for OpenAI-dependent builders.
-
Anthropic May Raise $50B at a $900B Valuation. Multiple preemptive offers have reportedly landed, with a board decision expected in May. Separately, Anthropic launched the public beta of Claude Security, powered by Opus 4.7, for enterprise codebase vulnerability scanning.
-
LIV Golf Loses Saudi Funding. Saudi Arabia’s Public Investment Fund confirmed it will stop backing LIV after the current season ends in August, leaving the league searching for a replacement backer and players potentially facing up to $90M in PGA Tour reentry fees.
-
Mayo Clinic AI Detects Pancreatic Cancer 3 Years Early. REDMOD reviewed nearly 2,000 routine CT scans originally read as normal and picked up 73% of eventual cancer cases. At two years before diagnosis, it spotted roughly 3x more early cancers than experienced radiologists did.
-
“Copy Fail” Linux Kernel Exploit Achieves Root on All Major Distros. CVE-2026-31431 is a critical logic bug present since 2017 that lets any unprivileged local user gain root via a 732-byte Python script. A patch is available; disabling the algif_aead module mitigates the risk. Discovery was AI-assisted.
Quick Hits
-
Waymo is increasingly blocking emergency responders in San Francisco and Austin, with fire chiefs reporting vehicles freeze up and obstruct access to stations. (Wired)
-
OpenAI announced it has already surpassed its 2029 Stargate goal of 10 GW of compute secured, adding 3 GW in the last three months alone.
-
Uber launched hotel booking via an Expedia partnership, with Uber One discounts and OpenAI-powered voice ride booking.
-
1X started full-scale production of its NEO humanoid robot at a new 58,000 sq ft facility in Hayward, CA, targeting 100,000 units by 2027.
-
Elon Musk’s SpaceX board approved a compensation package tying his pay to a $7.5T valuation and a permanent Mars settlement of at least 1 million residents.
-
Meta opened its ads platform to third-party AI tools via a new MCP server, letting advertisers manage campaigns through Claude, Cursor, or any connected agent.
-
Microsoft is rolling out Xbox Mode to all Windows 11 PCs, converting any Windows machine into a console-like gaming experience.
-
Two House committees opened probes into Cursor-maker Anysphere and Airbnb over their use of Chinese AI models (Kimi and Qwen).
-
Spotify is rolling out “Verified by Spotify” badges so listeners can distinguish human artists from AI-generated personas.
-
Eli Lilly Q1 revenue hit $19.8B (vs. $17.62B estimate), up 56%. Mounjaro +125%, Zepbound +80%. New weight-loss pill Foundayo is attracting 1,000 new patients daily.
Shower Thoughts
“You never experience yourself in the same way as others, but that version of yourself is the only one that exists in society.” “Wireless chargers use more wire than wired chargers.” From The Hustle