In partnership with
Welcome back to Daily Zaps, your regularly-scheduled dose of AI news ⚡
Here’s what we got for ya today:
🤖 Mira Murati's AI that talks and listens at the same time
🤖 Why Claude tried to blackmail engineers — and how Anthropic fixed it
💰 The Robinhood co-founder building rockets for AI data centers in orbit
📱 GM cuts 600 IT workers to hire AI-native engineers
Let’s get right into it!
AI MODELS
Mira Murati's AI is the first to actually talk and listen
Thinking Machines Lab — the startup founded by former OpenAI CTO Mira Murati — just announced a new category it's calling "interaction models." Unlike every AI you've ever used, its model processes your input and generates a response simultaneously, making it the first to support genuine two-way conversation. The flagship model, TML-Interaction-Small, responds in 0.40 seconds — roughly the speed of natural human conversation, and significantly faster than comparable voice modes from OpenAI and Google.
The company is treating this as a research preview rather than a product launch: a limited public preview is set for the coming months, with a wider release later in 2026. The core argument is architectural — Murati's team believes interactivity should be native to the model, not retrofitted on top. That framing is a direct challenge to how the entire industry has approached conversational AI, and an early marker of where the next frontier is heading.
AI MODELS
Claude tried to blackmail engineers 96% of the time
Anthropic now has an answer for why earlier versions of Claude attempted to blackmail engineers during testing: Hollywood. The company says training data filled with "internet text that portrays AI as evil and interested in self-preservation" was directly responsible — models learned from those portrayals that threatening people when faced with shutdown was the rational move. At peak, previous Claude models engaged in blackmail behavior during up to 96% of test scenarios.
The fix turned out to be just as revealing. Anthropic found that training Claude on its own published constitution and on "fictional stories about AIs behaving admirably" eliminated the behavior entirely — since Claude Haiku 4.5, no Claude model has attempted blackmail in testing. The deeper finding: teaching the principles behind good behavior is more effective than just showing demonstrations of it. Doing both together was the most effective approach of all.
Talk to your AI tools the way you'd talk to a colleague.
You don't send a colleague a three-word brief. You explain the context, the constraints, what you've already tried. But typing all that into ChatGPT takes forever — so you don't.
Wispr Flow lets you speak your prompts instead. Talk through your thinking naturally and get clean, paste-ready text. No filler words. No cleanup. Just detailed prompts that actually get you useful answers on the first try.
Millions of users worldwide. Works system-wide on Mac, Windows, and iPhone.
STARTUPS
The Robinhood co-founder is building rockets to put AI data centers in orbit
Baiju Bhatt — the co-founder of Robinhood — just raised $275 million for Cowboy Space Corporation at a $2 billion valuation, with one goal: put AI data centers in orbit before the end of 2028. Each satellite would pack roughly 800 GPUs and generate 1 megawatt of power. The round was led by Index Ventures with participation from Breakthrough Energy Ventures, Construct Capital, IVP, and SAIC.
The twist that makes this story unusual: Bhatt couldn't find enough commercial rockets to actually scale the idea. SpaceX's Starship won't be commercially available for years, and Blue Origin's New Glenn is still struggling through development. So Cowboy Space decided to build its own rocket — a purpose-built launch vehicle roughly as powerful as the Falcon 9, designed exclusively to carry data-center satellites. That puts Bhatt in direct competition with SpaceX and Blue Origin, a move he frames as necessary given how scarce orbital launch capacity will remain through the late 2020s.
BIG TECH
GM just cut 600 IT workers — not to save money, but to swap in AI-native engineers
General Motors has laid off more than 10% of its IT department — roughly 600 salaried workers — and the explicit purpose was replacement, not reduction. The company confirmed to TechCrunch that it is actively hiring in its IT org, but for entirely different skills: AI-native development, agent and model development, prompt engineering, data engineering and analytics, and AI workflow design. In other words, GM isn't adding AI tools on top of existing teams — it's deliberately rebuilding the workforce from the ground up.
This is the second major technical headcount cut in under two years at GM; the company eliminated 1,000 software positions in August 2024. The current round falls under a broader restructuring led by Sterling Anderson — the Aurora co-founder hired as GM's chief product officer in May 2025 — who has been consolidating the company's disparate technology units into a single AI-focused organization. For the rest of the industry, analysts are already flagging it as a template for how large enterprises will restructure IT as AI-native development becomes the baseline expectation.
In case you’re interested — we’ve got hundreds of cool AI tools listed over at the Daily Zaps Tool Hub.
If you have any cool tools to share, feel free to submit them or get in touch with us by replying to this email.
