I bet @KentonVarda wasn’t thinking about running a full browser engine on Cloudflare Workers when he created them in 2017. Kitesurf, our stateless browser for AI agents, is now open source.
https://t.co/0B1Xw1jqnm
Matt's right (again!)
~6 months ago, I learned agents needed a lot of constraints to successfully work in a "home repo" (https://t.co/TnWnhDhFZB)
But agents are so good now that you can set it up with 1 prompt and no special tools:
https://t.co/tCHS2et7cx
Working with agents across multiple repos is a pain
Where do you put the skills? The steering files? The tooling?
Let me propose a solution I've been using for months (that I thought was too boring and obvious to post)
It's a pattern called the Home Repo. Enjoy:
@joelhooks@samgoodwin89 That intro quote was basically me October 2025. Sam was right. 🫡
Alchemy v2 is my goto example for how clean Effect SDKs can be.
Switched harnesses and didn’t realize my “never comment on my behalf” rule was codex-only.
Sorry for the PRs with Opus 5.5 call diffs, summaries, evidence, and 2 way doors 🙃
Built a multiplayer AI harness for my family to use.
• shared sessions
• separate access/keys/limits
• family skills anyone can add (raise the floor)
It's a great way to help others steer and learn together.
All built on @Cloudflare.
My non-🌶️ take is that mono/many-repo decisions are easier to make than the org or tooling decisions behind them.
imo agents don't make monorepos an antipattern – they've just turned existing friction into a 🔥:
• Shipping the org chart
• Every change subjected to the same CI
• Bottlenecked by 1 release process
These are all solve(d/able), but the effort is usually higher to improve it for an org than to unburden a single team.
ime the sweet spot is to optimize for https://t.co/rXrYlA3Wes so 1 feature = 1 shippable unit
Big Walk is now getting review-bombed because they disclosed that they asked an AI chatbot how to fix a bug and incorporated the 10-line suggested fix into their codebase. https://t.co/eaDNHIuwXT
Today (Nov 10), ~55% of the steam reviews are negative, whereas previously they were 95% positive.
What the hell?
I just can't even fathom what's going on in the head of someone who is upset that a developer asked a chatbot a question and used the answer, so much so that they wish failure on such an amazing and wholesome game. But apparently there are an enormous number of people who feel this way?
How are we going to reconcile this?
"Materialize new views as you click "intents", or upsert visual changes in your existing interface. Software starts to feel magical."
Now we can add "fake doors" to product UIs and let the model generate the rest 👨🔬
Introducing an RFC for intent-driven replay-able generative UI.
Generate an interface. Click a button that intends to take you to a new page, generate that page on the fly. Coming back to that page is instant.
https://t.co/oqF8l9Rkli
Couple of interesting topics to dive into:
Infinite Software
Rather than hitting dead ends in your favorite software tool with wishes it could do X, Y or Z... what if you could just tell it to add the functionality and it appears inline with the rest of the content. You get to mutate and mold the experience to your expectations. Mix capabilities from your favorite tools (e.g. Add a "create ticket" button next to a list of Github repos that when clicked automatically create a Linear task).
Intents
Today software is usually made as a whole, meaning every path needs to be well thought out before we can ship it. Most users don't use 100% of the surface area either and everyone seems to use it differently. With intents you can place a button on the screen with an intent, such as "Open a new page with details about this musical artist", and without having to have generated that view ever before we can ship the software. When a user clicks that button that screen will be generated on the fly!
Materialize new views as you click "intents", or upsert visual changes in your existing interface. Software starts to feel magical.
Persistence
An important component of this RFC is the ability to persist changes over reloads, and consistently rebuild the entire UI. Page reloads reference the JSON page structure to lay it out. Tool calls and API requests are made on load to populate latest data into the slots in the UI – but we only make the calls we need. LLM calls stay totally out of the loop on every refresh, only used to amend experiences.
Data Connections
Bring any MCP server or OAuth service as a configuration and access their data. Write operations require human approval, read operations drive the visible UI generation. Ideal world would be you set up all your connections one time and use them anywhere you go. Github, Linear, Cloudflare, any service... you name it.
UI Library Agnostic
The RFC does not require use of any specific UI library so you're encouraged to bring your own. Bind specific UI components to a catalog. For example I bind "Modal" to "LayerDialog" from Kumo (https://t.co/hVxa6xUzlB)
----
Imagine not being limited by the developers vision and being able to expand upon existing products however you want. Add new functionality on the go. Today the concept works on green field projects, but will continue exploring what a product might look like if this option was given to users to mutate their UI.
The RFC specifies Intents and Persistence (IAP): a JSON document format and the host behavior that lets a model-generated application be stored, reloaded and replayed predictably.
Give it a read and let me know what you think!
I've decided to start working on Foldkit full-time. :)
There's enough momentum behind the project that I feel confident taking the leap.
No matter how things work out, I think I'll be glad I gave this my complete attention right now.
I'm excited to see what full-time focus does for the project and community.
More in thread!
imo these 2 affect business' ai productivity the most:
Fragmentarion - every local setup prevents raising the floor
Weakest link - all gains are throttled by the most constrained component (spending caps, avail models, MCP access, Slack v gChat integration, shared harness)
@woutstns Avoided sending a PR for exactly this reason - don't want to add to the noise.
Maybe it's better to get buy in on a PR elsewhere (here, discord) and the maintainer can have their agent PR the fork themselves.
I dunno...
Opus 5.5 > Sol 6.1
Maybe I'm late, but I couldn't get Sol to nail down why passkeys weren't working in T3 Code.
Opus showed me how there's an Electron limitation that Codex Desktop doesn't run into because it ships a web_authn.node shim.
Then it patched the nightly with a fork