TOP LAUNCHES

AI is talking behind your back

• ClayHog tracks how brands show up across ChatGPT, Gemini, Perplexity, Claude, and AI Overviews. It shows where you are mentioned, how you are framed, which competitors get pulled in instead, and what sources are actually driving those answers.Code, Codex, and other MCP clients can generate and modify design assets that are linked to your design system.

• Stagewise is a coding agent that lives in the browser and actually sees what you are building. It reads the DOM and console, lets you click on elements in your app, and then edits the code behind it directly instead of guessing from a prompt. It runs locally, works with existing codebases, and keeps you in control of the setup.

• Chrome Skills lets you save your favorite AI prompts and reuse them across the web. Instead of retyping the same thing over and over, you turn it into a shortcut and run it on any page with a click.

• Claude Code Routines let you save a prompt, connect it to a repo, and have it run automatically. You can trigger it on a schedule, from an API call, or off something like a GitHub event, so it keeps doing the work without needing to be reopened.

• Elytro is a self-custodial crypto wallet built for agents, not just humans. It lets AI agents send transactions, manage funds, and interact with Ethereum — but within rules you define, like spending limits and approvals, all enforced on-chain.

SPONSORED BY

DigitalOcean Deploy 

DigitalOcean Deploy is a one-day event focused on what happens after the model works. Latency, throughput, cost per token, reliability — all the things that show up the second something hits production.

There will be a fireside chat with NVIDIA’s Kari Briski on agentic AI. Teams from Character, Workato, VAST Data, Arcee, and the vLLM ecosystem will be there breaking down how they are actually running inference at scale plus a first look at what DigitalOcean is building next.

April 28 in San Francisco. Free to attend.

WHAT’S HOT

Did Anthropic just build Skynet?

Probably not, but last month they introduced, or should I say gently announced Mythos, a model it is not releasing to the public because it thinks the thing is too dangerous.

Why? Because this is not just a better chatbot. Anthropic says Mythos can operate more like a high-end software engineer, finding serious vulnerabilities and figuring out how to exploit them at a level that starts to make people nervous.

That is the real story here. Not the benchmark scores. Not the model name. The fact that a major AI lab looked at its own model and basically said: yeah, maybe not.

Experts are split on how alarming that is. Some see it as a big jump in offensive cyber capability. Others see it as the same trend continuing, just with a scarier label on it.

Either way, Mythos feels like a line-crossing moment. The era of ship first, patch the fallout later is starting to look a lot less comfortable. Anyway, Anthropic are still shipping tons every week, so go check out their latest launches!