Leon Dror Heuser (@LeonDHeuser) profile picture

Founder of @ReviewForest - Help the planet. Get more reviews. ⭐🌳

Berlin
Joined April 2009
Photo by Leon Dror Heuser (@LeonDHeuser)Photo by Leon Dror Heuser (@LeonDHeuser)Photo by Leon Dror Heuser (@LeonDHeuser)Photo by Leon Dror Heuser (@LeonDHeuser)Photo by Leon Dror Heuser (@LeonDHeuser)Photo by Leon Dror Heuser (@LeonDHeuser)Photo by Leon Dror Heuser (@LeonDHeuser)Photo by Leon Dror Heuser (@LeonDHeuser)Photo by Leon Dror Heuser (@LeonDHeuser)Photo by Leon Dror Heuser (@LeonDHeuser)
Claude Code kept maxing out my MacBook Pro. So I built a plugin that offloads the heavy work to throwaway Hetzner servers. Same idea as Vercel's /offload-to-sandbox, but about 20x cheaper for busy builds: 16 cores and 32 GB for €0.05 an hour. 🧡
At @vercel we got to test this model before release. Best way to summarize it: astonishingly brilliant. One interesting takeaway: it's very clear that this new generation of models can take on unprecedented bodies of parallel work. To actually seize this intelligence, you need cloud sandboxes. Full stop. The model can delegate and orchestrate far more than a single computer can handle. Likely far more than you can even imagine it can handle. Your bottleneck is now β‘  your ambition and β‘‘ your parallel computing power. Btw, I learned from @cramforce you don't need to overcomplicate it. Install Vercel CLI, setup the Plugin and tell the agent to use /πš˜πšπšπš•πš˜πšŠπš-𝚝𝚘-πšœπšŠπš—πšπš‹πš˜πš‘ skill. It'll provision and de-allocate cloud sandboxes as needed. You don't need a "multi-agent remote harness ralph orchestrator" πŸ˜†
3
2
81
Hetzner has no auto-expiry. Forget a server and it bills until you notice. So: β€’ every server gets a TTL and a watchdog on your Mac β€’ a launchd job sweeps every 15 min and after reboots β€’ it only touches its own labelled servers β€’ secrets are scanned out before every upload
1
19
My grandmother got out of Germany in time. Her family kept writing until most were murdered. I can't read the old German script, let alone find my way through an archive. Claude Code and I read every letter and wrote to 41 archives and institutions in 9 countries. Here's how ↓
1
2
60
The result: a memorial website our family in Israel, Germany and the US can read in their own languages. In autumn 2027, 13 Stolpersteine, brass memorial stones set into the pavement, are to be laid in Leipzig for the family: nine for the murdered, four for the survivors.
1
2
25
Have you tried researching your family's history with Claude? @claudeai @AnthropicAI
12
Tried this recipe on a real question: when do the next frontier models ship? Claude Code wrote the 2:26 video as code, @HyperFrames_ rendered it. First silent cut: about 40 minutes. Stills: GPT Image. Voice and music: ElevenLabs. Four Kling 4.0 background loops: $9.40.
Kurzgesagt has 25M subs. Millions of views on EVERY single video 🀯 nobody copies them for a reason: one video takes 1,200 hours and a studio of 50+ animators. we rebuilt their format with Opus 5.5 + @HyperFrames_ Studio. 100% code. a full 11-minute explainer. made in ~2 hours. under $50. want the template? we’re open-sourcing it for free πŸ‘‡
2
2
131
Anthropic and OpenAI ship on a rhythm: Opus every ~60 days, OpenAI's flagship every ~52, and each usually answers the other within a week. My Oct 28 call: Fable mid-to-late Nov, Opus by mid-Dec, Astra in Dec. Now "Mythos 5.5" and "gpt-6.2-astra" are leaking. Who ships first?
1
2
173
Two wildcards: OpenAI paused frontier training for 11 days and shelved GPT-6.1 Astra, and Anthropic's IPO roadshow can start Nov 10. Small samples, so give every date Β±2–3 weeks.
22