Twitxy Saved

Nico Bailon

Verified account This account is verified. Learn more

@nicopreme

Senior Agent Engineer @ Pika Labs / Pi coding agent early core contributor
Vancouver, British Columbia github.com/nicobailon Joined February 2014
1.7K Following 9.2K Followers 1.7K Posts

Latest posts 5

Pinned post

Nico Bailon Verified account @nicopreme
Created an agent skill called “Visual Explainer” + set of complementary slash commands aimed to reduce my cognitive debt so the agent can explain complex things as rich HTML pages. The skill includes reference templates and a CSS pattern library so output stays consistently well-designed. Much easier for me to digest than squinting at walls of terminal text.

github.com/nicobailon/vis…
0:47
119 replies 425 reposts 5.6K likes 1.5M views View on X
Nico Bailon Verified account @nicopreme
New Pi extension for improving your prompts: pi-better.

Type /better in front of your prompt and a forked subagent reads your code in the background, then rewrites the prompt with details you didn't think to ask for. Your agent can use it too when prompting its own subagents via pi-subagents.

Built on @felixrieseberg's prompt from felixrieseberg.com/how-to-be-more… (prob my fav prompt of the year)

pi install npm:pi-better
github.com/nicobailon/pi-…
1:17
43 replies 40 reposts 892 likes 31K views View on X
Nico Bailon Verified account @nicopreme
In my spare time I've been jamming on my own orchestrator: a native SwiftUI Mac app built on pi-durable (of course!)

I just talk to one "chief of staff" agent. It delegates async to an agent lead per project, each runs its own subagents, they all talk to each other, and only the decisions that need me land in my Inbox.

Because it hands everything off, the chief is always free: I can keep piling on new tasks across projects mid-flight and it keeps them all moving in parallel.

It renders in-chat gen UI built from real native Swift components, Mermaid diagrams via a native Swift port of Pele + rich HTML inline.

Fun as hell to keep iterating! Thanks @badlogicgames 🫡
54 replies 12 reposts 453 likes 27K views View on X
Nico Bailon Verified account @nicopreme
Hot take: Pi has extensions because I wanted an undo button.

Last December I opened a PR for Pi to snapshot every turn so jumping backwards in a convo could restore your file changes. Mario said no to adding it to core, then floated: "hooks could be a way to actually extend pi."

I opened the first hooks PR for Pi the next day. Mario rebuilt it into core based on my proposal, crediting me for the design and implementation.

I began shipping many Pi hooks right away, starting with pi-rewind-hook. A few weeks later, Pi hooks grew into a full-blown extension system. I kept contributing to core and building on top of Pi: 37 extensions and 23k+ GitHub stars combined.

Not bad for a rejected PR!
Marc Krenn Verified account @marc_krenn Dec 26, 2025
@badlogicgames @manuelmaly @nicopreme @aliouftw Just had a quick look at github.com/nicobailon/pi-… and this is almost exactly what I had in mind – great work btw @nicopreme! Should prolly work with some small adaptations, eg. substacking may need to emit a new event to trigger a rebase.
View quoted post on X
12 replies 7 reposts 275 likes 40K views View on X
Nico Bailon Verified account @nicopreme
Your agent skill was probably tuned for frontier models. On cheaper models they quietly fall behind.

New in skill-diet: skill-model-gap keeps your skill cooking on any model. It finds why one struggles and keeps only fixes that don't hurt the others.

github.com/nicobailon/ski…
0:44
Nico Bailon Verified account @nicopreme Oct 27
I've been dogfooding "skill-diet", a skill for slimming down agent skills and keeping them correct.

most skills load far more text than a task needs, and some parts often go stale. it also gets worse over time as AI updates skills.

skill-diet cuts what the model already does without being told, moves rarely needed detail into files it only reads when needed, and fixes what no longer matches the tool. then it runs the same tasks with the old skill, the new one and no skill, and keeps the new one only if agents do at least as well.

on my own skills: 70–80% fewer tokens, same results.

github.com/nicobailon/ski…
View quoted post on X