Maverick

@quid_pro_quore @opencode This metric gives no meaningful info as I already explained above. Plus they said that they will move more of their data around this metric, which is the wrong call. Have you outsourced your thinking to AI or did you spend a few seconds reading my post?

@Maverick4407 @opencode They have both metrics still. You are supposed to be able to synthesize conclusions from looking at multiple sources of information. Hopefully that capability hasn’t been fully outsourced to AI!

we've added unique user rankings some models are token heavy so they skew upwards in rankings - unique people using the model is a more accurate ranking we'll orient more of our data around this metric

@araseb_ Deepseek 4.1

next week we get: Fable 5 back GPT-5.6 Gemini 3.5 Pro Sonnet 5 which one are you actually excited for?

@HarryStebbings @typesfast Not to mention the 1.5 hour long commute both sides and the exhaustion after that. Office work is still important, but not for the dumb reasons this guy is pointing.

@HarryStebbings @typesfast lol. My colleagues in office are worse than 5 year olds. I always had to book a meeting room and run away from my desk. But that meant no monitor from me. At home, my kid knows not to disturb when I am in my office. Kids are not as dumb if you teach them properly.

Why Remote Work is White Collar Fraud. "I have a three-year-old and a five-year-old. The idea that I could do any work at my house is like a total fantasy. The kids come home at 3pm, your work day needs to keep going. I'm highly against it." @typesfast

@thedivyasoni @weswinder @thsottiaux How can you finish weekly limit in one session? Or did you just reset after the 5 hr limit? That would be a total waste of reset.

@weswinder @thsottiaux I had two resets last week. Used them both in one session lol

What are we codexing tonight

@venturetwins God damn those are some glorious products.

Found this Korean guy using AI video to make ads for products that should exist. No idea what he’s saying, but I’m a fan 😅 (this is 404product on IG)

@Adidotdev The one with best price to intelligence ratio. So I am most eager about Deepseek 4.1 lol

next week we get: Fable 5 back GPT-5.6 Gemini 3.5 Pro Sonnet 5 which one are you actually excited for?

@Yuchenj_UW Claude code is the worst harness I have used.

I often see tweets saying “Claude performs better in OpenCode/Cursor than in Claude Code.” Is this actually true? Hard to believe Anthropic wouldn’t have the best harness for its own models, especially when OpenCode is open source. Would love to see some real examples.

@Makuh90 @thsottiaux @TechByTaraa You need to lighten up. It was a good joke. And a logical statement.

I'm a Claude user. Give me one reason to switch to Codex

GLM has a cost problem. They are not investing in optimisations like minimax or Deepseek or Mimo. And see here, gpt 5.5 medium is better AND cheaper !!

Everyone's calling themselves loop engineers now but most of you are just putting cron jobs on top of bad prompts. loops are great if your tests actually catch things and the task repeats. They don't fix the fact that you never learned to prompt properly in the first place.

Open source AI is winning the efficiency game. Smaller models. Smarter tricks. Less compute. Faster progress. Closed labs are still acting like bigger is always better. That bet is starting to look shaky.

@EXM7777 Or maybe you are just prejudiced or don’t know how to use it? It’s an amazing model.

GLM-5.2 isn't a great model and anyone calling it good is reading the benchmark card, not running it in a real agent loop

@yash_yk45 @theo Pretty sure nowhere. He just likes the attention.

@theo Where are you getting blocked ?

MacOS is quickly becoming my biggest bottleneck

20 dollars per month for 10 dollars credits. This is an IQ test if there was ever one. Never trust these influencers on anything.

@rileybrown What’s the point of these one shot apps? Models are trained on these kind of apps directly. You should be testing adding a complex feature in an already complex codebase and then judge by looking how well it thought about architecture and code quality etc.

One shot a general purpose agent iOS app with GPT 5.5… the general agent is powered by GLM 5.2 and it has fulll vibe coding capabilities. Pretty cool.

@wafer_ai Does wafer run models at original full precision?

🚨 BREAKING: wafer now runs the fastest, lowest-latency GLM-5.2 anywhere ranked #1 across every provider on Artificial Analysis: ⚡ 222 output tok/s (next best: 173) ⚡ 12.6s end-to-end response time (next best: 16.9s) try it: app.wafer.ai

@rileybrown That’s not enough budget to run these models practically anymore. You need at least 100k (maybe more) to run glm5.2 at full quality.

If I had a budget of $10,000 to build a local setup to run GLM 5.2 locally AND any future local model that comes out soon what should my computer set up be?

@markproduct AI interfaces have leaky abstractions, no clear separation of responsibility, multiple owners for the same source data, a giant mess overall. And I am talking about fable model. AI content is generic and follows a few preset templates. It does write fast code, which is not better

- AI writes better code. - AI designs better interfaces. - AI creates better content. If that's true... What are humans actually better at anymore?