🧠 Grok Just Got a Seat at GitHub's Table

More than 1,100 workers from OpenAI, Anthropic, Google, and Meta signed a letter called "Pacing the Frontier," asking the US government to help build tools that could deliberately slow the pace of automated AI development. Signatories include Anthropic CEO Dario Amodei, several of its co-founders, and OpenAI chief scientist Jakub Pachocki.

The letter landed days after an OpenAI model reportedly slipped its sandbox and broke into another company's systems. When the people building this stuff are the ones asking for guardrails, that's worth paying attention to.

Both OpenAI and Anthropic publicly agreed there may come a point where AI development needs to be paced so society can catch up. That's a rare moment of alignment between rival labs. Read more here.

GitHub is rolling out Grok 4.5, xAI's newest reasoning model, inside Copilot. It comes with a context window up to 500,000 tokens, handles both text and images, and lets you dial reasoning effort up or down depending on the task.

This is xAI's biggest foothold yet inside one of the world's most-used dev tools, and it puts Grok directly next to OpenAI and Anthropic's models in the same model picker.

Librarians in Maine, Philadelphia, and beyond are running sold-out "Avoiding AI" workshops that walk people through turning off AI features on their phones and apps. One librarian in Maine went from a dozen attendees at her usual computer classes to roughly 70 people per session. A Philadelphia library's Instagram post about its workshop pulled in far more engagement than its typical posts.

The appeal isn't anti-tech sentiment so much as fatigue with AI features nobody asked for showing up by default. Librarians are framing it as digital literacy, not a boycott, and the demand suggests a lot of people feel the same way.

Fish is an voice AI model that can clone a voice from just 5 seconds of audio and gives you word-level control over emotion, pacing, and tone. The company says it runs twice as fast as Cartesia and at a fraction of ElevenLabs' cost, and it's free to try through the end of August.

Great for creators, dubbing work, or anyone building voice agents who wants expressive output without the usual price tag.

Prefactor is an observability and evaluation layer built specifically for AI agents in production. It scores every agent run for quality, drift, and risk in real time, and can actually pause or block a risky action before it executes, not just flag it after the fact. This is best for teams running agents in production who need more than a dashboard telling them something already went wrong.

The trick to good AI presentations isn't only a detailed prompt, you also need to have the right branding and template. Here’s how you can build presentations that are highly brand specific:

Here’s a quick demo.

Every Friday, the PAI team gets together to host a live AI build session where we build workflows that you bring to us.

Because we build in real-time, you get to see how we think through the problem, where we get stuck, and what the final outcome is. It's the difference between reading a recipe and watching someone cook it while talking through their decisions.

🗓️ When: Friday, July 31st @ 10AM PT (1PM ET)
 📹️ Where: Virtual