The kind of creativity (and brute force) agents showed to execute code when their sandbox was supposed to prevent them in the Hugging Face attack is bonkers. Seems like we're about to (repeatedly) learn a rather bitter lesson.

Dogmatic idealists comfortable in their cheap rent controlled apartments unwittingly conspire with rich NIMBYs in homes that keep appreciating while their property tax doesn’t, resulting in pushing out people replaced by new rich and keeping other newcomers out

We're building and growing quickly. Come join us 🚀

If you're looking to work on an incredibly ambitious product with a super talented team full of people you'll actually like, you couldn't do better than @nuance_ai.

AI will reconcile quantum and standard physics and still struggle with Zoom AV settings. This is the final benchmark.

Vague posting about @spc

Another insane Jev use case! Jev is making it dramatically cheaper to evaluate what actually happened inside an agent run. And finally, someone open-sourced a self-improving memory layer that can put that signal to work across agent harnesses: - Claude Code - Codex - Cursor - OpenCode, and 20+ more Beacon by @asymptotelabs continuously captures your agent history across harnesses and uses Jev to identify which runs are actually worth learning from. It then turns the highest-signal workflows, corrections, and debugging patterns into reusable skills. GitHub repo: github.com/Asymptote-Labs/ag… (don’t forget to star it ⭐ ) Beacon preserves the complete session history. But preserving a run and learning from it are two different things. Most coding-agent sessions contain routine exploration, failed commands, and fixes that only apply to one task. The trace can remain available for inspection without turning every detail into guidance for future agents. Jev scores each run for evidence, reuse potential, and human correction signals. An application policy then decides whether to promote, review, or discard it. The recording shows this in action. Claude receives a coding task, modifies the implementation, and runs the tests. I then provide an edge-case correction, so Claude updates the code and adds regression coverage. Beacon automatically captures the complete session. Jev evaluates whether the correction contains a reusable engineering lesson. Once approved, that lesson becomes available to other coding agents working on the project. Since it works across harnesses: - Claude Code sessions can teach Codex. - Cursor debugging can improve OpenCode. So a problem solved by one agent should not need to be learned from scratch by another. If you want to dive deeper into Jev, I also wrote a hands-on guide to building this Jev-style decision path with open models, entirely locally. Read it below.

The amount of new information that came out since @GoodfireAI co-founder @banburismus_ and I talked about the Hugging Face incident is....a lot. We spoke the day after the news first broke. Still a super interesting conversation!

Goodfire's Chief Scientist @banburismus_ talked about reward hacking with @JPBrebner last month at SPC: piped.video/watch?v=YoRuuz2Z…

Important work from @GoodfireAI ⬇

This is some 12 Monkeys stuff

What would it take to make AI writing less grating?

The pace and stakes of AI make me nostalgic for the silliness of web3. Remember NFTs? Remember when we were arguing about the value of jpegs instead of human extinction or transcendence? Simpler times.

Why you’re more likely to find something worth building in a group than alone.

When @adityaag and I interviewed @thejamescad for @SPC we got off the call and immediately said something like "that guy just really wants to win." I'd say @profound has proved that read out.

we’ve joined @figma! since we started this journey together, @zan2434 and I have believed that the most important challenge of our era is creating the human interface for increasingly powerful machine intelligence, while making sure it amplifies human creativity, agency, and judgment. we found that same conviction when chatting with the figma team. we’re excited to help build the human layer for creating and collaborating with AI, and we’re looking forward to sharing more soon.

Big news everyone - we’re excited to share that @cmmnknwledge, the team behind Flipbook, are joining @figma to help build the future of creating with AI. It is so rare to find a team and product at this scale that believes as much as we do in both the unlimited capability of AI and human creativity. The future is bright and we’re excited to share what we build together soon!

Excited to share @profound has raised a $180M Series D, co-led by @sequoia and @kleinerperkins. Winning in AI Search requires an inhuman amount of work. That is why we are building the AI platform for marketers, underpinned by two things: 1/ Your AI Marketer, a marketing expert that proactively investigates your data, identifies opportunities, and then does the work while you stay in the loop. 2/ Context Manager, a living synthesis of your meetings, email threads, and brand data powering this new teammate. As your brand evolves, so does your AI Marketer. AI labs are pushing the frontier in fields like engineering, law, and finance. But marketing, one of the economy's most complex and consequential applications, has received far less attention. That’s why we’re expanding our applied AI research team. We’re bringing together research scientists, engineers, and IOI medalists to advance AI for marketing. Their focus: post-training models for marketing workflows, and building the world's first AI benchmark for real-world tasks in the field.

Ideas means nothing without execution. They reach no one, without right articulation. The proverbial devil, can be in the articulation. We're running a hands-on storytelling workshop in BLR, by SPC storyteller @OnaRoy88. All you need to bring, is an artifact we can transform together. A demo, pitch, landing page...anything representing your story. It can be WIP too. You'll get feedback & tips, and leave with an artifact that represents your story, wherever it's at right now. 20 seats only. Luma Link below.