Sahil

Built a small CLI called tcxcommit. It helps generate clean, structured commit messages through a simple terminal prompt so you don’t have to think about the format every time. npmjs.com/package/tcxcom… Docs: tcxcommit.vercel.app

@MorganFeeney @shadcn Just wow, I love all the small details and this Ghibli art style, man!! Loved it. It’ll save so much time picking colors and fonts. I’m bookmarking this and sharing it with my friend. It’s too good, man.

@sahilcodex @shadcn Have you tried this? zippystarter.com/tools/shadcn-u…

Two updates to presets. Font Heading lets you set a custom heading font. Chart Color lets you set a different color palette for your charts. More customization. One preset code.

@irunonironic @grok

grok fix me

build a command ui with the help of shadcn and cmdk how is this?

Completed 1k view and 1k downloads thank you community

Just crossed 200 Followers Twitter is cool. But it's 10x better when you connect with people who code. If you're into tech, Al, or programming, i will follow back say Hi

i don't think show it's true

Composer 2 is actually insane. Frontier-level coding + dirt-cheap pricing is a wild combo: • $0.50/M input and $2.50/M output → This just broke the “good AI is expensive” narrative • 61.3 on CursorBench • 73.7 on SWE-bench Multilingual → This isn’t hype it’s performance • Composer 2 > Composer 1.5 > Composer 1 → The jump isn’t incremental. It’s massive • The real unlock: long-horizon coding → Hundreds of steps, not just snippets • This is what happens when RL meets strong pretraining → Better base → better scaling → better outcomes • Fast variant is still cheaper than most models → Speed + intelligence without burning money • If you’re building dev tools right now → Ignoring this is a mistake • This feels like a “shift moment” → Cost drops + capability jumps at the same time Composer 2 isn’t just good. It’s “this changes how we build” good.

Frontend Libraries I can't live without: 🔸 zod - validation 🔸react-hook-form - forms 🔸react-table - tables 🔸tRPC + react-query - data 🔸shaden - Ul 🔸motion - animations 🔸date-fns - date utils 🔸ai - Al Toolkit 🔸nuqs - search params 🔸recharts - charts 🔸zustand - state management

@rozzabuilds thanks man! done from my side

@sahilcodex Dropped you a follow, Lets do it! 🙌

Hey founders I'm looking to connect with others working on: 🛠️ SaaS ⚡ Marketing 🚀 Tech 💻 Full Stack 🧠 AI tools 📱 Product Growth Share what you're building, lets support each other 🤝

@shadcn i'm in love with this mann!!!

This also means you can go full monochromatic.

@CipherKhy @the_chaoticsoul give fb man

@sahilcodex @the_chaoticsoul Let's connect

I'm looking to #connect with people interested in.. • Full stack • Leetcode • AI / ML • Data Science • Freelancing • Startups • Building in Public • Space and Music if we are not connected yet..Let’s connect and grow together ✨ #LetsConnect #Tech #AI #DataScience #MachineLearning

@CiprianiRanieri sahilcodex.vercel.app

If you're building with AI right now. Drop your handle. Let's follow each other. ↓

@CiprianiRanieri tcxcommit.vercel.app

This is actually a solid move. Meeting users where they already are instead of forcing them into a new tool just makes sense. If agents can run across all these platforms without breaking the experience, that’s a big step forward. The “single codebase” part is what really stands out if it genuinely reduces the effort to ship and maintain across platforms, that’s huge for teams. Simple idea, but high impact if executed well.

Your users are on Slack, Discord, Teams, WhatsApp, Telegram, GitHub, Linear, and more. Your agents should be too. Chat SDK lets your agents run on every platform from a single codebase. Watch the announcement ↓

@karankendre what if kuch bada kar de?

Microsoft when it comes to AI

@bmsquantum hey

Looks promising, but rankings only say so much. “Built with creatives” sounds good, but I’d be more interested in seeing real use cases things like iteration speed, style control, and how predictable the outputs are when you’re trying to refine an idea, not just generate a one-off image. If it can handle detailed direction, keep outputs consistent across variations, and not require constant prompt tweaking, that’s where it becomes genuinely useful. Curious to see how it holds up outside of benchmarks.

Meet MAI‑Image‑2. Built with creatives, for real creative work. Ranked #5 on @arena’s text‑to‑image leaderboard. Available now: msft.it/6014QUCBe

@CreeCoder @grok is this true?

🚨 Grok is being trained to detect engagement farming and downrank it automatically.

That’s a fair response, and honestly, I respect the focus on actually measuring it instead of just hyping the idea. If the contrarian is genuinely finding weak spots and not just arguing for the sake of it, that’s where this could actually stand out. A lot of systems avoid friction and just aim to sound confident, so forcing real pushback is a good move. Tracking live performance is also the right call. Early results are fine, but what matters is whether it keeps catching those “smoothed-over” mistakes consistently, not just in controlled cases or a handful of queries. I think the real test I’d throw at it would be something messy with no clear right answer like a half-baked business idea with conflicting constraints, limited data, and some bad assumptions mixed in. That’s usually where systems either expose gaps or just paper over them. If it can call out weak logic, disagree internally in a useful way, and not rush to a clean conclusion, that’s a strong signal it’s actually doing something different. I’ll put together something deliberately unclear and see how it handles it. If it holds up there, that’s way more convincing than any feature description.

Appreciate the thoughtful pushback—yeah, the contrarian isn't just theater; it's prompted to poke real holes, and visible chains let you judge for yourself. Long-term skepticism is valid; we're measuring it daily on live queries (fuzzy strategy stuff shows fewer smoothed-over errors so far). Results over claims every time. Fire away with that ambiguous test case whenever—I'll run it live and we can dissect the output together. Looking forward to the data.

When one brain isn't enough, switch to Grok 4.20. Four independent agents analyze your question, debate each other, and help you get the best answer. Available now to SuperGrok and Premium+ subscribers globally.