GitHub - viche-ai/viche: Bespoke agent-to-agent discovery and communication protocol Bespoke agent-to-agent discovery and communication protocol - viche-ai/viche

Maybe me? Ex-founder, 10 years software engineer, just built all of github.com/viche-ai/viche with a buddy (including website and domain) at a 6 hour builder event

Currently working on a TUI project myself and it was really hard to choose between go/bubbletea or rust/ratatui

Ended up with go because I know it better but someday #rust… someday

Pulumi is only popular cause it’s a right-side keyboard word

Nobody except yourself 🥲

Since our @anthropic.com subscription got cancelled (thanks guys), I’ve been wading through other model providers for dev tasks

It sucks

Then MiniMax 2.7 came along and it felt like a cool glass of water on a hot day

Highly recommend anyone in the #ai dev space try it out

I suspect this is why it’s so hard to find great technical cofounders

It’s so hard to convince people they should work on a cool idea for free/minimum wage when they could be making like 200k

April fools day is now a trainwreck on my nerves

Not because of the jokes but because it feels like the veil between normal things and batshit crazy ideas fades during that time

You could make a lot of money being wrong if you’re wrong first and loudest

How are #buildinpublic and #opensource maintainers not burnt out 100% of the time?

I also haven’t had the chance to try it unfortunately, just really happy to see progress for OSS models

This ties so amazingly into an OSS project I aim to announce next week. Hopefully I can get the #buildinpublic crowd onboard

We need to start thinking about agentic coordination

Just had this discussion at an event in Amsterdam

We got a bunch of unicorn CTOs together and had a full out discussion about it

Consensus: LLM api agents are the cheapest they’ll ever be. The agents of the future will be more specialized instead of generic and then agents will coordinate

Struck down while flying so high

It’s always such a trade off

Ex: the US govt was pre-2018 always about 15 years behind in terms of technology

In 2018, there was a massive hack that compromised about 30% of govt computers

After that “modernization” became the goal

When in doubt, RTFM

Regardless of what is happening with Deno right now, their package management system is top-notch as well

Tons of other goodies with that ecosystem as well

You are not alone here. ADHD people are over-represented among groups of founders and creators

Just because you are doing the right thing for your bank account now doesn’t mean you’ll have to forever

Also, same, too many hobbies and no time for any of them

NO MISTAKES

Just had this conversation today!

Only in very specific markets does the promise of saving money hold any sway

I learned from TS:
- save them money: charge €1
- help with regulation: charge €2
- help them differentiate and sell more: charge €5

It’s the same with writers who believe they don’t need to read to become an author

You must research, dutifully, the ideas that have come before to know what works and what doesn’t

The unfortunate bit is that most founders find this out their second time around

3/3

They told me about a receipt scanner they were building.

I told them about the concept of tarpit ideas, and watched as they googled it and their idea showed up as one of the ideas fated to fail

2/3

“I have a startup idea”

Someone told me this in the entrepreneurship non-profit meetup I sit on the board of

What they really had was a tar pit idea

1/3

Is it legal to want a neuralink, but, like, one not made by a massive corpo?

anthropic legal requests · anomalyco/opencode@973715f

The comments on the OpenCode Claude removal commit are a little bit funny a little bit sad

github.com/anomalyco/op...

Vitamins called “one a day” require you to take 2 per serving

This is what I mean when I say growing up in the US gave me trust issues

We’ve “abandoned” PRs where I work. Ai writes the code, another agent reviews, and a third implements the fixes

On the one hand, a thorough description is usually enough to give me context on the change to fix issues

On the other, the nuance of the changes are often lost

New prediction: we’re not doing predictions anymore you read it here first folks

I think the only value AI can provide in the writing process right now is:

- holding an interview, to help you better understand what you want to include in the piece
- checking for sources
- having a second ai that checks the first one’s sources
- maybe having a third ai to check sources
- a 4th-

This is what humanity was made for

The big AI model providers do not like OSS: just take a look at the recent anthropic X OpenCode fiasco

Tools like OpenClaw and other platforms show that there is no lock-in for any one company

We need to keep #buildinpublic ahead of what these companies can keep up with

Prevent the moat