Adam Kerr

@jack @michaelneale This is awesome! Though I can't tell if this is the last season of silicon valley or the first of battlestar galactica...

mesh-llm: pool compute to run open models. built by @michaelneale at block: docs.anarchai.org

@PeterDiamandis Can we get a local AI XPRIZE?

Now tell the world — futurevisionxprize.com

In case you haven’t heard, we put up $3.5M+ to source optimistic visions of the future. We don’t care who you are or where you are. Tell me… what future do you want for your children?

@PeterDiamandis Decentralized AI, food and energy

Iykyk

@IAmSneyon @TheAhmadOsman This is true but temporary

@TheAhmadOsman @zero_to_seed Most frustration with local models isn’t the model it’s the setup around it. Performance is only one part of the equation.

if you’re using Ollama switch to llama.cpp if you’re using OpenClaw switch to Hermes these are basics at this point

@heygeorgekal Beauty vs utility

Why are people allergic to AI generated art... but perfectly fine using vibe coded apps?

@sandislonjsak Maybe they set up guardrails so that QA is less of a bottleneck? Though I’m with you I’ve maxed out at 3-4 due to QA

My brain simply can't run more than 3 agents in parallel and QA all of their work. I am sure I am not the only one. How do people manage 10 at once? Or they simply lie?

@cowincrisis moo

if you woke up one day and you were a cow what’s the first thing you’d do

@david_nix @TheAhmadOsman Yea too many holes for me that I didn’t want to manage, need to sleep at night 😅

@zero_to_seed @TheAhmadOsman Whey Hermes over OpenClaw? (The complexity of openclaw scares me to death for security.)

@aakashgupta The whole AI world is full of false binaries. I guess polarization gets more likes 🤷 you can see it in the AI patterns too (it’s this, not that). Two things can be true

A tweet went viral claiming Claude Code doesn't use evals. Bosses started pinging their teams asking if evals even mattered. People who had been championing evals at their companies suddenly had to defend the practice. Ankur Goyal's response: the tweet is factually incorrect. People at Anthropic are using Claude Code, providing feedback about whether it solves their problems, and incorporating that feedback into iterations. That is a form of eval. They don't have a formalized process with a third-party tool. They don't need one. The people training the models, building the harness, building the product, and using the product all sit inside one building. The distance between builder and user is zero. That's the part most people missed in the discourse. The question was never "does Claude Code use evals." The question is how much formal eval infrastructure does YOUR team need given YOUR distance from the end user. Anthropic engineers solving problems for other engineers at the company that trains the models can get away with structured vibe checks. A healthcare AI company where engineers aren't doctors and patients are worlds away from the codebase cannot. The tweet created a false binary. Evals or no evals. The real spectrum is distance. Zero distance means informal feedback loops work. Maximum distance means you need data sets, scoring functions, and production monitoring or you're shipping blind. Most AI teams are nowhere near zero distance. If you're reading this, you probably need more eval infrastructure than you have.

@thiojoe Build a binary only computer 😎

Anyone else think it's weird that computers take even multiple seconds to boot up? I feel like they should take milliseconds

@asaio87 Cheaper and smarter every day

Soon we will face it. AI is expensive. And dumb.

Frontier labs charging $200/month for subsidized LLM use with rate limits is like a drug dealer giving you the first bag for free.

Design the rules and the architecture and the pattern you want and put in checks to ensure they happened and then make sure they happened. That is the new coding. nobody writes binary anymore because higher level languages were invented. The quality of the lower level code still matters but nobody writes it.

Because these codebases need to architected and maintained You cannot maintain a large codebase with full confidence unless you yourself are in said codebase doing the work. The more you allow LLMs to code for you, the less you understand of your own codebase, and the less youre able to maintain in the long term.

@yacineMTB The demand for intelligence is insatiable. The more you can do the more you want to do.

There is so much that wasn't worth even attempting that has become cheap enough to do. So much work and attention that is worth doing. So many ideas that are now possible. Everything done creates more things to do. I never have made as many decisions as I do these days

It's remarkable how much of my work is completely automated w/ AI, and yet, I still am necessary. The amount of time I personally have to spend working just isn't going down. Instead, the leverage of my own time is going up. Every second I spend not working becomes more painful

@theo Why

Local models? Really cool. The local model community? Absolute cancer.

@icanvardar token spend needs to tie to business objectives. just because it’s easy doesn’t mean it’s useful

when did it become all about spending ai tokens, not building the project?

@npew Everything is markdown

Markdown is the new Python

@AiFreak_ @kenanatalay @TheAhmadOsman Best UI is no UI

@kenanatalay @TheAhmadOsman @zero_to_seed Lmstudio is llamacpp + frontend you don't (usualy) need :)

@AlexFinn How would you relate it to Hermes

Anyone who thinks AI is a bubble has never used OpenClaw