yc p26 | cofounder https://nitter.cf/t.co/K8w5kRIDbM | prev. cto https://nitter.cf/t.co/GTa4g3K7xX, swe & data eng @ https://nitter.cf/t.co/9PeGXxKtn4 | writing at https://nitter.cf/t.co/AfKvla6D1Y | p(doom) over 0.5
Paris, France
Joined January 2026
- Tweets504
- Following231
- Followers701
- Likes1.3K
very bullish on @s43stha
Today, We're Introducing Multipeer
→ Run multiple Claude Code, Codex, Grok, etc all at once on the same canvas
→ Collaborate peer-to-peer on the same WiFi, or over the internet (fully encrypted)
→ Jump into a teammate's terminal and take control
→ Combine multiple Claude accounts and tools of your own together
→ iPhone app (coming soon): run your agents from anywhere
Work used to be something we did together. Somewhere along the way we all ended up alone in our own tabs. We're fundamentally changing this!
Now that everyone is a builder, we need one interface for everything, powered by your agents and tools.
That's what we're building.
what can be more fun than a clever linguistics experiment?
> the same experiment with dogs
sciencealert.com/dogs-seem-t…
if only they had @superlogYC
Superlog now suggests automations
every day you get 3 suggested workflows
based on what integrations you connected and what issues you've had
find new performance wins every week, automatically.
here's the template we use at @superlogYC:
superlog.sh/automations/new?…
- checks your observability data
- finds slowest operations
- hillclimbs
- send you a PR
A new template at superlog.sh
Keep docs up-to-date: a weekly scan of your database that sends you a PR with anything you forgot to fix in your documentation.
superlog.sh/templates/update…
things we shipped today to superlog.sh:
- triggers on @github pull requests opened or merged
- integration with @devrev
- using @vercel as a context in your automations
🚀 never stop shipping
anthropic.com/research/self-…
i was skeptical of 'if anyone builds it everyone dies': if a model escapes, it will need massive inference from somewhere.
either from aws or ant / oai. at some point they will notice?
turns out nah
our clients asked for usage analytics - so we shipped them (proudly powered by @autumnpricing )
now you can see which automations churn the most tokens
thank you @johnyeo_ @autumnpricing ! surprisingly, this is vanilla js/css
(will probably move to echarts as we add more use cases to the component)
we just rolled out our @ChatGPT plugin for superlog.sh
the entire experience to set up your automations is inside the chat!
Been trying @superlogYC, if you're looking to automate agents tasks on your observability stack it's a game changer for us
Plugged on @axiomhq and it suggested awesome missing monitors for us
we made this video with opus 5.5
here's the workflow:
1. brainstorm the script / bullet points
2. manually create raw stills in Figma. quick ugly screenshots and approximate positioning
3. use opus 5.5 / astra + figma mcp to polish stills and extract UI from screenshots
1/n
you can use superlog.sh to triage alerts in Slack
- get a reaction 🟥 / 🟧 / 🟨 / ✅ to prioritize alerts without having to manage 10 Claude sessions separately
- Superlog will tell you which customers and orgs were impacted
- get a PR with a fix right away
1/n
4. use opus 5.5 and remotion to animate the stills
5. show it videos you like!
that's it!
the biggest productivity hack is:
- using LLMs for tedious work like drawing polygons and animations
- Figma/@paper to refine details and add your taste
