iiinova — where your digital self keeps evolving

Nimo 🏳️‍🌈 @nimobeeren.com · 28/08/2026

And also you could also use that domain knowledge to improve models in ways that big labs can’t

Nimo 🏳️‍🌈 @nimobeeren.com · 27/08/2026

What is this sketchy link? 🧐

110

Nimo 🏳️‍🌈 @nimobeeren.com · 26/08/2026

That's real. Maybe it's time for a different form of art? Or just different subjects :)

Nimo 🏳️‍🌈 @nimobeeren.com · 25/08/2026

Why do models still default to ThreadPoolExecutor over asyncio? Am I missing something?

000

Nimo 🏳️‍🌈 @nimobeeren.com · 23/08/2026

Having tried a large-scale migration (Vite + FastAPI → Next.js), it seems like models struggle to generate idiomatic code when given an existing implementation with very different idioms Wonder if it would work better to generate a spec and implement that without ever showing the old implementation

000

Nimo 🏳️‍🌈 @nimobeeren.com · 22/08/2026

A lot of AI tools feel kind of “top-down” in that they give you a whole bunch of stuff at once that you just have to trust to work well. I much prefer working in a bottom-up way, adding things only when I feel a concrete need and I feel confident in the way the tool satisfies that need.

Nimo 🏳️‍🌈 @nimobeeren.com · 18/08/2026

So it has no wings? Interesting 🧐

Nimo 🏳️‍🌈 @nimobeeren.com · 17/08/2026

Border radii are my passion.

A stack of MacOS windows, each having a slightly different border radius so they are all visible

000

Nimo 🏳️‍🌈 @nimobeeren.com · 16/08/2026

I’ve been trying to cherish those natural breakpoints. Sitting in a cafe and just looking around or out the window while I wait for some computation to happen.

Nimo 🏳️‍🌈 @nimobeeren.com · 16/08/2026

This is really cool! It keeps secrets out of agents hands by swapping them in at the network boundary. Even without a sandbox this is an upgrade since a lot of secret leakage seems to be agents accidentally reading files and they shouldn’t rather than intentional prompt injection.

190

I should not have posted this, the turn ended up taking 40 mins 😭

000

The True Cost of AI Coding

I'd skip the Fable hook at the start, but I really liked the rest of this video. It's worth being conscious of the way AI coding tools make it harder to stop when you need to. youtu.be/iPUn1Fnfn0k?...

youtu.be

YouTube video by Syntax

I guess prompting is mostly about figuring out how to transport thoughts to another entity. If you understand how you arrived at those thoughts, you understand how the agent could do the same. Hence, you need to explain less. Makes sense that thinking about your thinking makes you better at that!

000

Working on a side project where deadlines are of no concern, I actually kind of like waiting a few minutes for my agent to respond. Gives me the chance to compact the conversation in my head.

120

Just brainstorming: it could be a step towards AI companies demanding more compensation for specific uses or controlling what their model can or cannot be used for.

000

Love using domain models for this, e.g. using @mattpocock.com’s skills. The code tends to follow pretty naturally. And yeah LLMs really haven’t gotten much better at architecture over the past 6 months I feel.

Nimo 🏳️‍🌈 @nimobeeren.com · 11/08/2026

I think you learn so much more from consistently doing things “the right way”. That allows you to reach higher peak output quality which is much more rare and valuable.

000

Nimo 🏳️‍🌈 @nimobeeren.com · 07/08/2026

Great idea, and it strikes me as kind of odd that it doesn't exist yet. I do wonder how well the HTTP spec (which probably defines a message as a string following some kind of grammar) could be captured by a JS object. Though I suppose anything is better than a raw string which is status quo!

110

Nimo 🏳️‍🌈 @nimobeeren.com · 06/08/2026

Woah those stacked ones are so cool!

Nimo 🏳️‍🌈 @nimobeeren.com · 29/07/2026

I have a lot of web dev and TypeScript people in my feed, but I think I want to diversify into Python a bit more! Who should I follow?

120

Reposted by Nimo 🏳️‍🌈

Ken Jennings @kenjennings.bsky.social · 10/07/2026

When the weather gets hot, please take special care for the goths in your life. With their floor-length coats, platform shoes, and velvet capes, it’s hard for them to thermoregulate as we do.

13972211185

Nimo 🏳️‍🌈 @nimobeeren.com · 26/07/2026

For me personally, Pi isn’t the best fit. I’d rather have a team actively developing the harness over the flexibility of building my own. It’s a different target audience!

Swizec Teller @swizec.com · 22/07/2026

Reposted by Nimo 🏳️‍🌈

outsourcing: They do task, you think and worry delegation: They own outcome, you build feedback loops this is a note to myself

041

Nimo 🏳️‍🌈 @nimobeeren.com · 21/07/2026

I’m impressed by the polish of the OpenChamber iOS app (in TestFlight). So far it’s been even better than the desktop app! testflight.apple.com/join/5ek6GU1E

testflight.apple.com

Join the OpenChamber beta

Available on iOS

Nimo 🏳️‍🌈 @nimobeeren.com · 20/07/2026

I'm really looking forward to the next hillclimbing problem I have to solve.

001

Nimo 🏳️‍🌈 @nimobeeren.com · 19/07/2026

Are there any polished, open-source iOS apps for interacting with agents generally? Doesn't matter what harness

000

Nimo 🏳️‍🌈 @nimobeeren.com · 13/07/2026

A good thing about usage limits in coding tools is that it forces us to be aware of the different costs of models. It's still hard to pick the right model for the right job, but at least we have an incentive to do so now.

Nimo 🏳️‍🌈 @nimobeeren.com · 03/07/2026

I knew Fable was token-inefficient, but didn't expect it to use 400K tokens and 20 minutes just to create a PLAN for a migration. That's like 3x the total tokens of my entire codebase, including dependencies...

Nimo 🏳️‍🌈 @nimobeeren.com · 29/06/2026

Wait is this like a “prove you are an agent” challenge?

Nimo 🏳️‍🌈 @nimobeeren.com · 04/06/2026

Thanks for the ping! It’s a little short notice for me so I’ll have to pass on this one, but future invites are very welcome!