Yash Rathi

@SwiggyCares @Swiggy After 40 mins when the agent got connected they are offering a 20% refund. When 50% of my food is missing in WHAT world 20% refund makes sense!

I ordered my first order from toing app @SwiggyCares @Swiggy . I ordered a footlong subway. Got my order after 1 hour, and it is missing a sandwich. Have been trying to reach customer support for the past 40 mins and it's still unable to connect and it still didn't connect me!

Horrible service! Horrible customer support!

@rork Really want to try

Rork is the easiest way to use Claude Code with Opus 4.5 to build real mobile apps, publish to App Store in 3 clicks and start making revenue Reply to get a free 25$ subscription. Ends tomorrow!

@help_delhivery @help_delhivery @delhivery I have DMed you the tracking number but no one has responded. Kindly respond

@yrathi1 Hi Yash, We apologize for the unpleasant experience and any inconvenience caused. At Delhivery, we are committed to providing our customers with a seamless experience. Request you to DM us your tracking number for us to assist you better. twitter.com/messages/compo…

I booked a courier from my college to my home using @delhivery. They lost my package in transit that contained many valuables and important documents. Now I am only getting offered 1K compensation on that. Please help @help_delhivery @delhivery

@help_delhivery I have sent the DM

@davis7 gpt-5.2-high if you're on cursor, and 5.2 codex on Codex CLI. Codex models feels hypertuned for their own CLI

Since Claude's down I think I'm gonna try the GPT series. Any recs on which one to use? First thought was GPT-5.2 High, but what about GPT-5.2 Extra High Fast (latest)? Maybe GPT-5.2 Low Fast, that seems like a better idea than GPT-5.2 Low (latest). Maybe a Codex model, GPT-5.1 Codex Max High Fast is tempting, but perhaps I should consider GPT-5.1 Codex Extra High Fast? There's also GPT-5.1 Codex Max Medium Fast? Oh but wait, there's GPT-5.1 Codex Max Extra High Fast (I almost missed out on the extra level of high on my GPT-5.1 Codex Max model glad I caught that). There's also GPT-5.2 Pro which I could use to write a plan for GPT-5.1 Codex Mini to go implement, the question there is should I use GPT-5.1 Codex Mini Low (latest) or GPT-5.1 Codex Mini High? There is also GPT-5.1 Codex Low which could be faster, but would it be faster then GPT-5.1 Codex High Fast? It does have Fast in the name after all...

@wassgha @vercel @davidhdev This actually looks very cool! Great work!

Finally got around to updating my personal site after 5 years. A fresh new portfolio, built on @vercel and @davidhdev's React Bits. → gharbi.me

@alxfazio I have definitely been thinking of the same. There's something like this with OpenSkills

mcps and skills badly need a package manager. whoever builds this first is gonna make bank

@huang_chao4969 This looks super clean

Just open-sourced AnyTool ✨ - tackling some real pain points in agent tool orchestration 🛠️ Building AI agents is still pretty challenging - even with MCP around, tool orchestration has become one of the biggest headaches. We're seeing issues like overwhelming tool contexts, unreliable community tools, and limited capability coverage. AnyTool tries to solve these with a universal tool orchestration layer that lets AI agents interact with virtually any tool more reliably - from web APIs to system commands to desktop automation. 🔗 GitHub: github.com/HKUDS/AnyTool Key features of AnyTool: ⚡ Fast tool retrieval - Multi-stage filtering to quickly find the right tools from any toolkit 📈 Self-evolving tool orchestration intelligence - System remembers which tools work well and continuously improves selection 🌍 Universal coverage - Supports any MCP servers, any system commands, any GUI automation, and more One line of code to give any agent smart orchestration with any tool. Basically turns messy multi-tool workflows into a single API call. Welcome to try it out and share feedback! #AnyTool #AIAgents #ToolOrchestration

A 7M parameter model just outperformed Gemini 2.5 Pro, and o3-mini on reasoning. Tiny Recursive Model (TRM) is 10,000x smaller but achieves higher accuracy on ARC-AGI & Sudoku. - Costs <$500 to train (2 days on H100s) - Runs for <$0.01 per task

@VanpeltVentures Marketing is a pretty major part of building anyways!

Marketing isn’t hard It’s just not as fun as building

@aidenybai This looks cool! could you send the link as well?

you can now try React Grab's agent to directly edit the website no signup / use for free

@JulianGoldieSEO 100%. My current stack is Stitch for ideas/iteration -> Cursor for the heavy lifting. The pipeline is seamless.

Everyone is sleeping on Google Stitch. Here’s why that’s a huge mistake → → It redesigns your entire website from a screenshot → It writes usable HTML and CSS → It builds UI components, not flat images → It exports directly into AI Studio → It lets you iterate with simple natural language The entire design → code pipeline just collapsed. Save this video, you’ll upgrade your whole stack. Want the SOP? DM me.

@yeasindesign 6 months fully booked on one thing sounds like deep work heaven (and maybe a little bit of hell haha). glad you enjoyed it, that's a huge achievement.

I’ve never shared this before. But I wanted to share a small story about the biggest project I had taken once. The project was one of the largest we’ve ever worked on. A product built from scratch, valued close to a billion dollars. The whole team didn’t just work on it; we lived it. The energy, the pressure, the excitement… it was unlike anything we had experienced before. We were fully booked for the next 6 months on this one project. Enjoyed every moment of it though, through and through.

@startupideaspod honestly the 'ship one feature' part is where i always mess up. i try to build the whole thing at once lol. needed to hear this.

Want your SaaS to blow up fast? Keep the launch dead simple. Loom video, one feature, posted on YouTube. * LocalRank: $5K MRR on day one * YouTube: drove 80% of sales * Enterprise pricing: 4x'd revenue overnight Ship one feature. Let YouTube do the selling. YouTube had 1/10th the views of their X post but drove 80% of sales. Trust converts better than reach. (we like that) Here's the 4 step playbook you can steal: Step 1: Ship One Feature Fast Don't build everything. LocalRank launched with just a heat map. Test the market before you build the product. Step 2: Stack Your Distribution They launched with: * X thread * YouTube video * Email list You don't need 100K subscribers. Jackie had ~10K when he launched. Step 3: Cap Early Users Artificially constrain supply. Creates urgency. Lets you raise prices. LocalRank raised prices 4x doing this. Step 4: Add Enterprise Packages Some customers have 10x the budget. Base plan: $40/mo Scale plan: $400/mo Top tier: $1,600/mo Revenue jumped from $5K → $20K MRR. Start with one feature. Post a Loom to YouTube. Keep Shipping and Posting You got this

GitHub (Star it!): github.com/farouqaldori/c… Website: claudeisland.com

I just tried Claude Island by @FarouqAldori and I really like it! It is a 100% open source native app that lets you control Claude Code directly from your MacBook's notch. Link is below ↓