Md. Abu Taher

@nathan_tulu I got confused as well, then I tried other harness, like pi or omp. And over there the models are mentioned properly.

What is this, suddenly I don't have access to Composer 2.5 through Grok Build? Only Grok 4.5 available.

So many different providers. Codex, Claude, Grok, Kimi, Qwen, Cursor, GLM. Everyone is offering something valuable. But you have only one pocket.

@kinder_grinder @ctatedev @unsplash This looks impressive, is this the new extension?

@ctatedev @unsplash

For the first time ever, agent-browser, portless and json-render have made it into a model's training dataset 🥹

@ZixuanLi_ My GLM sub expired two days ago. The limits were horrible for new subs.

This might be the best time yet to subscribe to the GLM Coding Plan 🔜

It's interesting all of their cat has this exact same figures. I tried a few times and the result had the exact same kind of fish in the wall, the ears are same, body as well. What's more interesting is, almost all of the models including sol max produces same/similar color for the cat as if there's cats in this color in the world.

@_0xpainn I am using a paid cursor subscription and the results are great so far!

SpaceXai made Grok 4.5 FREE in Cursor 😳 it's their top coding model with 500K context. 4x fewer tokens than Opus on coding. it doesnt need API key or credit card. Just one toggle and you get frontier coding. what you get for $0: -83.3% on terminal-bench 2.1 -64.7% on swe-bench pro -500K context for big repos -4x more token-efficient than Opus 4.8 what this replaces: -Claude Opus 4.8 -any paid model credits you're burning in Cursor how to enable (10 seconds): Cursor → Settings → Models → toggle Grok 4.5 Important: Limited time offer. No API key needed. Just a toggle. Your buddy pays $20/M for Opus output in Cursor. You get Grok 4.5 for $0. use it while it’s still available

I am not impressed by kimi k3 yet. The service is not capable of handling the requests yet, and it still talks too much. Not sure how people are getting great results. I gave it this simple prompt: create an SVG of a cat drinking milk in morning close to a window and a boy is seeing him outside and wrap svg in html. save it as kimi.html It went on rambling for 40 minutes, before timing out, another run took 1:30 hours before timing out again. Their interference is under pressure for sure, doesn't explain the rambling though. Qwen finished it in 8 minutes. Twice. Output are identical to each other, fish in the frame, cats ears etc. I gave the same prompt to Luna, Sol Medium, Sol High, Sol Max, GLM 5.2 - Short (a faster variant by NeuralWatt). Most of them finished within 2-4 minutes. Strange enough the artifacts created by Qwen max preview are quite same no matter how many times we try. It's possible we need to provide a more specific prompt, add some restrictions and guidance. Oh I see why they call it Sol Max. It's extremely detailed.

None of the four. For me it was just shipping the smallest possible version and telling users straight up that perfecting it would take forever. Taking action moved the product and moved me.

Builders, what could actually moved your product forward? - A viral post - A useful integration - A community - Something nobody expected

@neerajjj6785 I'll disable signups, I should be satisfied 1k people are looking at my products. Kimi did it yesterday. They have so many users they stopped onboarding new customers.

Interviewer: Your API works perfectly with 1K users. It falls apart at 100K. You can't change the infrastructure. What would you use to fix it?

@DanielSmidstrup @mynameisyahia has been recently to YC for @getcontextdev, he might be able to provide some guidance.

Y Combinator's next application deadline is July 27. I've been fascinated by YC for years. But now I'm not sure it's my game anymore. I'm building in public, not chasing a perfect product. Is applying still worth it when your path no longer looks like YC's?

My friend @thesturag pushed me a ton for this since the last month. We got a basic mvp ready last month, that's why it was possible to redo this again within a few days.

Alhamdulillah! We got out first annual customer on a vibe coded product. Most credit goes to Grok and Luna. It took many iterations, lots of tokens from GPT, Grok, Kimi and GLM to get this done, oh and lots of coffee. Grok was specially helpful today, way more than Luna did yesterday. This is doable! Let's do it!

But none of my replies are reaching anyone 😅, what a weird situation.

@CreeCoder @questguyontrip This account is from 2014, for last few days, no matter where I replied my replies were not visible. It would simply not show up. Then I got a warning notification, and then another that they didn't find anything wrong with my account, so they continued my account.

@questguyontrip So it wasn’t all bots but suspensions?

X did another bot 🤖 purge. I lost ~100 followers instantly. How many did you lose?

@jun_song I use GLM 5.2 Fast from a provider and it's been good so far. Waiting for them to pick up K3 soon.

Open Source AI updates : > GLM-5.2 ✅ Almost Opus-level > Kimi-K3 ✅ Almost Fable-level > Qwen-3.8 🔜 2.4T, Expected to beat Opus > Deepseek V4 GA 🔜 $0.0028/M Expected to beat Opus > Minimax-M3-Pro 🔜 3T, Expected to be Fable-level > GLM-5.5 🔜 Expected to beat Opus Holy

@trikcode People has my data already anyway, what else do they want?

switched my Claude API to Kimi yesterday it’s faster, 10x cheaper, and writes better code. i don't care if Beijing knows my db schema. it's not like my startup has users anyway.

@ammar_nassri I think suddenly becoming reply guy will also trigger the algorithm and you will get issues down the road.

Does this mean I'm officially out of the reply guy phase? I know the reply guy never dies, but it's good to know we're moving forward

I think I got shadowbanned two days ago. 😅 So suddenly becoming the active reply guy is also an issue. Algorithm hits you real hard. Now none of my replies are even visible anymore.😂 Gotta continue what I gotta continue.

@Clintonkadry I think I got shadowbanned. 😅

Getting shadowbanned can destroy your business on X I’ve avoided it for 8 months Here’s how: I don’t spam comments I don’t use offensive words I don’t use automation to send DMs I don’t mass follow and unfollow people I wait at least 3 seconds between each like That’s it Stop doing things that make your account look like a bot

My cousin was using one PoS solution for $100/yr for one of his small shop. Recently the subscription expired and he asked me for help. I decided to build one. And after a few nights of work, I got a fully working prototype ready. ChatGPT, Cursor, Grok played a big role here. This is still mvp but the new era is coming. Soon big subscriptions for small shops will be replaced by small hand or ai crafted solutions because it's just easy to do this way.