Sebastian

swe @ salesforce, tech lead for agentic cost optimization

[Sebastian @SebyTremblay]

@bercankilic shift

[Sebastian @SebyTremblay]

scrolling on twitter is the human form of continuous pre-training

[Rishi Alluri @RishiAlluri]

It’s finally here: real-time voice hallucination correction.

Exciting to turn a race against the audio playhead into something reliable on live calls.

Check out the article for more https://t.co/LuIBWA8YA0

[kache @yacineMTB]

you can outsource your thinking
but you cannot outsource your understanding

[Sebastian @SebyTremblay]

@nikhil_suresh2 Super cool stuff! Physical AI 🆙

[Sebastian @SebyTremblay]

@RishiAlluri Never ending

[Sebastian @SebyTremblay]

https://t.co/CQeWm4Een3

[Sebastian @SebyTremblay]

Finally, it helps keep track of your agents. When shipping large features, PRs can quickly sprawl into thousands of lines of code.

This method helps me mentally model everything the agent did and have confidence I am shipping high quality code without sacrificing velocity.

[Sebastian @SebyTremblay]

I started using GitHub as my agent control plane.

This unlocks:
1. A compounding dataset for agent skills
2. Transparency into my work (for correction/teaching)
3. Increased alignment/supervision on large changes

Here's what I love about it 👇

[Sebastian @SebyTremblay]

It also gives rich telemetry into your work and how you think about problems.

Teammates can learn from your workflow and step in to correct you when you drift off best practices.

[Sebastian @SebyTremblay]

https://t.co/VTXLh1OtJj

[Sebastian @SebyTremblay]

https://t.co/LDCpY4Y518