Christoph Sitter

Last week's @RaiseSummit was on fire, and not just because of the Paris heat. With my background in AI and software, it was good to meet model providers, infrastructure companies, and application builders pushing the ecosystem forward. On the infrastructure side, we had direct conversations about powered land and data center development, the space we work in every day at @fiftysixtyhq The event was packed, so getting into talks took time, but the queues ended up being just as valuable, leading to some great conversations and new connections. The after parties, timed with the France Morocco match, were a fitting way to close the week. Huge congrats to @LandoInvests and @HenriDelahaye for an amazing event. Looking forward to the next one!

Looking forward @RaiseSummit

@chadbyers Congrats on the Venture! Would be very keen to learn more and see if we can work together. We have some interesting sites with near-term power availability that may be of interest to you.

We recently funded a new neocloud that has unique access to hardware. We're building in an ecosystem that has been largely ignored, but is massive. We're hiring across all engineering roles, but specifically for folks that have helped stand up and scale data centers. Reach out directly if you want to learn more.

It all comes down to vetted schedules for the complex, overlapping workstreams that make power delivery real. Through exclusivity at sites and ranches that provide access to near-term power across the US, that is exactly what we’re building at @FiftySixtyHQ.

Today, @editframe emerges from stealth. Agents need video. Editframe Agent Skills: npm create @editframe@latest Just prompt Claude Code, Cursor, or Codex and get a working video or a full interactive GUI. This video was created just by prompting 👇

@PeterDiamandis Reach out if you’d like to discuss how we at @FiftySixtyHQ navigate the core challenges causing these issues, along with our near-term powered land and data center projects 🤠

Some people like beach photos. We like this.

Biggest mistake: I treated memory like storage, not like accumulated experience.

A few weeks ago we gave an AI agent a real job: Data center developer. It ran diligence, sent emails and worked on live projects. It was shockingly good. Then I accidentally wiped its memory. Turns out: when AI systems learn, infra mistakes don’t just break things, they erase experience. We're now building it again, but better. Full story 👇

@christoph.sitter/we-gave-an-ai-agent-a-real-job-then-i-accidentally-wiped-its-memory-7d543f87dad7" target="_blank" rel="nofollow noopener">medium.com/@christoph.sit…

This is what the protocol should look like everywhere. Claude Code gave us the hooks. We built the integration. Zero-touch persistent memory. Now imagine if ChatGPT, Gemini, Cursor all had the same hooks. One config. Every tool remembers you. That's the future. memobase.ai

Here's what Memobase + Claude Code HTTP hooks does: Session starts: memory capabilities auto-injected. No system prompt. While you work: tool usage silently tracked. Before compaction: session activity preserved. Session ends: everything persisted for next time. Paste one config. That's it. Memory just works.

The problem with AI memory: there's no easy way to share it across tools You can build the best memory layer in the world, but if the AI tool doesn't call it at session start, users have to manually configure system prompts. Claude Code just shipped HTTP hooks. And we use them to solve it.

Memobase is live in open beta. The universal memory layer works today. The open protocol is the mission. If you're tired of being the memory layer yourself — try it: → memobase.ai

The harder part: getting AI tools to actually USE memory automatically. There's no standard for ambient memory — no protocol that lets you plug in your memory server and have it work everywhere without manual setup. That's the mission. Build that protocol.

I've been building a fix for #aiamnesia — one memory layer that works across all your #AgenticAI tools. Here's the problem, what's live, and why this is harder than it looks. 🧵