Maybe: Christian Rocha

Conversation is 50% listening and 40% discerning what not to say.

I wish they called them Earth, Wind and Fire

@ivanburazin Local models, let’s go

Someone on our team wanted to copy text from a screenshot. He uploaded the image to Claude and asked it to retype the text. I watched him burn tokens to do what Preview on Mac does natively for free in two secs. And this is happening everywhere. Everyone is overusing inference for every little thing because it works, and the cost feels abstract. OCR, file parsing, simple lookups, structured data extraction, and the like. Even agents are constantly using inference for tasks that should be deterministic. Once the cost pressure catches up, the stack will change (more cpu workloads, fewer gpu workloads)

@tadericson @charmcli Dude, let’s go! If you need some inspiration: github.com/emprcl/sektron github.com/emprcl/signls

how do we @charmcli @meowgorithm this project

Day 1 of a new @grok #grokbuild project, this time >> to make a #node based Music collaboration tool #colab #daw #audio #musiceditor #digitalaudioworkstation #indieartist repo ~ github.com/fornevercollec… live test at ~ fornevercollective.github.io/Qbpm/

No developer actually says ‘call sites’

Sometimes I write some code to fix a bug then puff up my chest and brag to @andreynering He usually informs me that a better fix is in a PR and ready to go. These PRs are typically authored by him. This is how I came to understand the phrase "seeing around corners."

GLM 5.2 is now on Hyper. 1M context window. Zero data retention. MIT-licensed. GLM 5.2's incredibly capable. This is a moment.

@MelkeyDev I didn’t know Vercel was hiring male models

Big send in London

We’re doing a pink-yellow deployment here at @charmcli…please stand by

@oanaolt Let’s not go that far

I guess AI makes NYC just like SF

@iamdothash @bsideup @charmcli Nice! Feel free to ping if you have any questions about any of our libs. You’re building some great stuff!

@meowgorithm @bsideup @charmcli I can't type some days. It is updated to v2 today!

I've used Kubernetes for years, and it was time to make the TUI suitable for me. Inspired by k9s and lens. Here is kli.

@bsideup @iamdothash @charmcli I beg to differ! Charm is indeed a great showcase of Italian power, akin to Florence and Venice at the peak of the Renaissance!

@iamdothash @charmcli Wow, my dyslexia hit hard here 😂 “Great showcase of its power” 🤦‍♂️

@iamdothash @bsideup @charmcli I must ask: why are you using v1? There’s so much to gain in v2, most importantly rendering performance.

@bsideup @charmcli It's the best, isn't it. Using it for all of my TUI.

@ClementDelangue Charm, most likely

I’m in NYC for the day. What are the best AI startups or people here?

We’ve upgraded Kimi K2.6 on Hyper. More capacity + better caching + faster responses. Zero data retention, always.

Or as we’ve been saying internally, CheapSeek

Now I see the full picture

Qwen 3.7 Max is now on Hyper. We waited a week to make sure it met our standards for performance and zero data retention. This one is impressive. Try it here: ⚡️ hyper.charm.land

@tweetatweir Same. Honestly I miss coding Elm quite a lot.

@meowgorithm Looks🔥 Looking forward to spending some time with it

Elm is not always the easiest language to program in, but it might be the easiest to maintain, and I love it. And now some madman has created a byte code compiler - fully native, no node or JS anywhere. @meowgorithm discourse.elm-lang.org/t/run-native-e…

@kwuchu Oh very nice

At work, I found myself drowning in tabs when landing features because there are different pages for CI, merge queues, and deployment statuses across services, it's quite a fragmented experience. Being the terminal junky I am, my latest vibe-coded side quest at work is a tool I built named rollout. This tool lets folks monitor all of their pull requests, including CI build status, merge queue, and deployments to dependent services. Additionally, if there are any failures in any of these steps, there are error messages that the user can copy and paste into their agent to easily resolve it as well. Also, you can use various subcommands like `rollout pr pr_num` or `rollout --user username` to see the status of your teammates' PRs and if they have landed if you depend on them, in addition to integrating into `tuicr` for a code review experience, all in the terminal. (Highly recommend checking tuicr out btw, it's good software). This easily has become my favorite tool that I use at work, it just makes for a great experience, especially when coding with agents. I shared it internally on Slack and other engineers are loving using it, somebody even told me today, "i have too many github pages with all my PR's open and i hate it, this is exactly what i need, and its not something i even knew i needed lol" Completely built the MVP of this using Go, BubbleTea, and Codex 5.5 Medium while waiting for a deploy, happy with the end result for an hour or so of prompting.