Gabriel Garrett

The secret to success: they WON'T tell you: > sit down > become very still > try to do absolutely nothing for as long as possible > keep doing that until nirvana is reached > escape the endless cycle of death and rebirth

GLM-OCR 0.9B model that beats Gemini on OCR benchmarks going live in 15 minutes to test it on real-world datasets and build some cool demos link: youtube.com/watch?v=_KCXD8…

imagine getting harpooned to death by this thing. a more dignifying, yet infinitely slower and more painful ending than an explosive drone

Jensen Huang: "If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed. This is no different than a chip designer who says 'I'm just going to use paper and pencil. I don't think I'm going to need any CAD tools.'"

MPP comes with a directory of 100+ services for your agent to use & pay for autonomously on Tempo: • @alchemy — blockchain data across 100+ chains • @alliumlabs — onchain analytics • @browserbase — headless browsers • @dune — onchain SQL queries • @fal — image, video, and audio generation • @merit_systems — phone calls, travel, email, & more • @p0 — web search and deep research • @postalform — mail delivery via USPS • @prospectbutcher — sandwich ordering • ... and many more No API keys. No accounts. mpp.dev/services

Some of you are too busy doomscrolling to notice that borderline singularity level ML has been achieved in OpenSource Software Models. The Minimax-2.7 model blog describes how the agent literally runs and iterates training on its own. You could be a researcher who wants to prototype 4 different ideas but you don't know if they will all work. You run 4 of them in 4 parallel worktrees and ask the agent to write the code and parallelly run the training runs and then pick the best ideas that work. During the runs M2.7 discoveres effective optimizations for the model: systematically searching for the optimal combination of sampling parameters such as temperature, frequency penalty, and presence penalty etc. and this helps you pick the best ideas fast and improve while basically only making critical decisions instead of handwriting all the code and infra. This means that the model is "almost" self evolving and with a few improvements to infra+harness+model engineering we are right at the footsteps of exponential improvement. I say this with no exaggeration this is going to be wild year.

During the iteration process, we also realized that the model's ability to recursively evolve its harness is equally critical. Our internal harness autonomously collects feedback, builds evaluation sets for internal tasks, and based on this continuously iterates on its own architecture, skills/MCP implementation, and memory mechanisms to complete tasks better and more efficiently.

Not enough people talk about how unpleasant vibecoding is. The best analogy I can think of is driving. It's cool that we can just hop in a car and drive to the store. It's a lot faster than walking. And yet, it's so stressful and infuriating, we had to invent a new word just to describe its effect on people: "road rage." AI-assisted coding is the same. It's so much faster--there's no going back to coding everything by hand, the equivalent of walking everywhere. And yet it's incredibly annoying and stressful. It's characterized by annoying delays between requests, time-wasting misunderstandings, blatant lying, and absurd overconfidence. Hopefully this gets better as models improve.

wow. just got a debit card with MPP -- took <30 seconds.

@craigzLiszt been thinking it’s not a bad idea lately given the rhetoric

started learning hebrew, just in case

@Austen Stripe provides this

Surely someone has built this: Where’s the API that is connected to all of the AI providers and lets me charge per token with a markup?

@0xyoussea I don’t think the bazaar is set up well yet tbh. You can’t even search anywhere as a human to see if a broadcast you’ve made for a merchant worked.

These services seem to only be discoverable for humans, while x402 includes the Bazaar, a discovery layer for finding agent-ready APIs. Conversely, MPP has vertical integrations with the network, wallet etc, while x402 relies on ecosystem solutions (CDP, Base, Solana etc)

An honest comparison of MPP vs. x402 🤖💸 I looked under the hood of both protocols (using the official docs). Both leverage HTTP 402 for payment requests, but their architecture, positioning, and execution are totally different. Here’s the technical breakdown 🧵👇

agents can send physical mail for you now via mpp thru services like @postalform through @stripe 👀 cool to see this type of real-world payments orchestration in action

only bottleneck is consuming code, so make sure to tell codex that you want just that: "write extremely easy to consume code, optimize for how easy the code is to read. make the code skimmable. avoid cleverness. use early returns."

@mayfer Did you use parakeet v2 or v3? I only ever use v2 because it’s an English only model and it’s pretty much perfect for me

PSA, apple's built in Speech Recognition is both faster and more accurate than Parakeet. whisper-large-v3-turbo is still king for accuracy, but if you need something faster just use apple's built in. you can use on-device only if you like. use github.com/AppDevGuy/OSSS… to integrate into native apps. other transcription models like Qwen or Mistral etc are not good for short utterences at all. they only shine with longer context

it's not AGI when it's looking at a 10k line file and at no instance ever says "this file is getting too long, let's break it down into easily manageable components"

@levelsio use screen

I hate tmux It's so incredibly user unfriendly The shortcuts make no sense I wish someone would make a better tmux Even just logging into tmux attaching the screen is an illogical hell to type Again I hate tmux, it's so shit

@bubbleboi well, it helps that the image isn’t real

I don’t even understand how flames can get that big

@wholemars where’s 14.3 though

In a few months people will be kicking themselves for not seeing what was obviously coming

@Nexuist counterpoint: the intelligent vibecoders are not flagrantly bragging about it

It should fill you with peace that everyone vibecoding everything is losing money instead of making it and the only exceptions to this rule are people who were making money before vibecoding too