Aviv Sheriff

@ImGordonSun fixed

@Avivsh cant DM

Introducing Simmy: the Youtube for playable stories. Today, Simmy is #3 in its US App Store category, starting with the $1B romantic fiction vertical. We believe playable stories will redefine entertainment forever. Comment for an INVITE CODE to our public beta. (1/8) THREAD 🧵

I'm a single-issue (local) voter. Leaf blowers.

@saranormous Sure

any nontechnical folks want to get more comfortable/powerful in their use of AI and want to be a beta user on something I made?

@aphysicist thanks!!

@Avivsh let's goooooo, excited to try it

Built the open source Starcraft APM dashboard for vibe coding. pip install motif-cli && motif live

Also inspired by @idosal1 and his work on RTS interfaces for agents

@andrewchen Accept all changes and use tests and good hygiene instead. I.e. regular refactoring, guidelines on hygiene, formatted, etc.

One question I've been asking founders is: do you try to review all the code that the LLMs write or do you just accept it? I think it's about 50-50 right now but the momentum is towards just accepting the AI-generated code and I think that number will eventually go to 100% This is one of the most telling indications of how AI-native a team is. It's hard to get super high throughput if you are reviewing every line Poll: what do you do?

@garrytan I've run into the cookie problem before trying to automate scraping and literally copy pasted from dev tools :D. Does it only work on Mac?

Like this is pretty cool being able to pull over cookies from your real browsers - it makes the headless browsing much more useful

Comic book guy on Product Hunt can never win

@CameronSorsby @AlphaSchoolATX Love it

We’re launching a new @alphaschoolatx high school for aspiring entrepreneurs. Our promise: Make $1m by graduation, or receive a full tuition refund. Yes, this will be the coolest high school in the world. And we're building the best team in the world to make it happen. We’re looking for 2-3 exceptional coaches to help us guide the students towards achieving this aggressive but achievable goal. You won’t be giving lectures or assigning homework. You’ll be grilling them on their P&L, driving them to the car wash they bought, critiquing their email funnels, pushing them to do things 99% of the world doesn't believe is possible. Job posting is live and DMs are open.

@bcherny @openingai_com Does it work in cli?

Update: this is now rolled out to 100% of users

@alex_prompter V nice. But also seem most valuable for enterprises or companies hwere one skill is used thousands of times, such that the cost of skill "training" pays off. Not sure that ill train every skill in my Claude Code for 7%

🚨 R.I.P. making AI agent skills manually. Sentient and Virginia Tech just built a system that discovers and refines them automatically through failure analysis alone. It’s called EvoSkill, and the numbers speak for themselves: - Claude Code on a brutal 89,000-page Treasury document benchmark: 60.6% → 67.9% after just 1.5 epochs on 10% of training data. - On SealQA (designed to trip agents up with noisy, conflicting web results): 26.6% → 38.7%. Here’s how it works: Three agents run in a loop. - An Executor attempts tasks. - A Proposer diagnoses every failure. - A Skill-Builder materializes a fix into a structured, reusable skill folder. Only skills that improve held-out validation scores survive. The model stays frozen the entire time. No fine-tuning. No retraining. But here’s the part nobody’s talking about: Skills evolved on one task transfer zero-shot to completely different tasks. A search-persistence skill built on SealQA was dropped unchanged into BrowseComp. Accuracy jumped from 43.5% to 48.8% with zero modifications. That’s the gap between prompt optimization (overfits to one task) and skill optimization (captures general capabilities). The future of coding agents isn’t bigger models. It’s agents that learn from their own failures and build reusable expertise.

@marclou nice work. love this project!

I just built Google Search Console integration for TrustMRR. Now it verifies: 🔄 MRR 👀 Visitors 💰 Revenue 📉 Churn rate 🖱️ Google search clicks (NEW) 🔍 Google search impressions (NEW) No self-reported metrics. Everything is pulled from APIs. Should I rename it TrustEVERYTHING? P.S. Startups listed listed for sale with analytics integration get 5% ranking boost for being transparent.

@marclou Human slop indistinguishable from AI slop?

Apple is vibe coding too

@amasad I wonder if vibe coding on Replit/Lovable/v0 will catch up to Cursor/Claude Code where CLI/IDE will no longer be superior for 99% of use-cases.

“While 12 tasks were building in the background, I was experimenting with design variations in canvas view at the same time.” 🤯

I wonder if Microsoft hits back? I assume that they want to own the AI stack on Excel and PPT or do they view themselves as a complement? MSFT can probably see that this is where the wind is blowing. MSFT has a similar distribution advantage in business that Google has in consumer. If they release their own Gemini-tier AI that works better with Excel+PPT, they can prevent that market share from Anthropic.

Claude for all things knowledge work feels like it is on a very similar trajectory to what agentic coding experienced last year. I expect entire industries that rely on spreadsheets and powerpoints to begin to be transformed in the next few months.

@trq212 Nice touch. I used to always have a chatGPT browser open for this purpose.

We just added /btw to Claude Code! Use it to have side chain conversations while Claude is working.

@unnamed1tw @melvynx Sorry I meant usage limits.

@Avivsh @melvynx Cursor doesn’t have rate limits as it would not be great to have to stop working when you are in the flow.

Just so people know: I used Cursor for 4 days with API credits enabled and spent $536 This is the REAL cost of coding with AI Claude Code and Codex are just hiding it If VC money stops, we'll all be paying $200 a day just to code with frontier models

@rauchg don't you still need a dedicated sim/phone number for it to operate independently? or whatsapp business aPI? that's the friction, because I don't want it to have access to my history.

1️⃣ 𝚗𝚙𝚖 𝚒 𝚌𝚑𝚊𝚝 2️⃣ 𝚌𝚛𝚎𝚊𝚝𝚎𝚆𝚑𝚊𝚝𝚜𝙰𝚙𝚙𝙰𝚍𝚊𝚙𝚝𝚎𝚛() 3️⃣ There's no step 3 Easiest way to start building WhatsApp agents