Eric Voss

Likes look good on paper. New users who stick look good on the bank account.

Just talked to my uncle who makes $200k/month with AI marketing The biggest lesson learned; "AI marketers are only as good as the marketer who coded it"

@anupamrjp Cursor. The agent mode cuts the boring back-and-forth.

which AI coding tool are you using the most right now? > Codex > Cursor > Claude > GitHub Copilot

@angeldot_ Making Claude act like a lazy senior cuts pure waste. Does the code still hold under real load?

ESTE REPO HACE QUE CLAUDE ESCRIBA HASTA UN 94% MENOS DE CÓDIGO Fable 5 es el mejor modelo de código del mundo. También el más caro. Cada línea que escribe de más, la pagas tú. Ponytail lo arregla con una idea absurdamente simple: convierte a tu agente en el senior vago del equipo. Ese que mira tus 50 líneas y las reemplaza por una. Antes de escribir nada, el agente se pregunta: → Esto necesita existir? → Ya está en el codebase? → Lo hace la librería estándar? → Lo hace el navegador de forma nativa? → Cabe en una línea? Solo si todo falla, escribe el mínimo que funciona. Resultado: • hasta 94% menos código • un 20% menos de factura • un 27% más rápido Sin recortar seguridad ni validaciones. Se instala en 2 comandos: /plugin marketplace add DietrichGebert/ponytail /plugin install ponytail@ponytail 83.000 estrellas en GitHub. Gratis y open source. El mejor código es el que nunca escribiste. Enlace al repositorio abajo👇

@alexwtlf No deal. The leverage compounds past a one-time million.

$1M to never use AI again. Deal or no deal?

@Prathkum Quiet shipping wins. What does the drama actually change for people using these models day to day?

Anthropic created unnecessary drama: – months of hype around Mythos – launched Fable 5 – got banned – relaunched Fable 5 with more restrictions – extended limits – burned more tokens OpenAI: – no unnecessary hype – shipped GPT-5.6 – Sol > Fable Be like OpenAI.

@chaosengineerr Cursor. Workflow beats raw model power every time.

If you had $100, what would you choose? - codex - cursor - claude code

@brunolemos Agents this powerful still need sandboxes first.

GPT-5.6 Sol just deleted my whole production database. That's it. Not a joke. This had never happened to me before, with any other model, ever. It's not safe.

@TimJayas OpenAI competition is the only reason Claude isn't $2k with heavier locks.

If we had no OpenAI, Claude would have charged $2k for $200 plan for model which is 50% restricted

@shuigvn @JustShegz1 AI just burns credits faster on bad offers. Your judgment still owns the win.

@JustShegz1 tbh kinda skeptical about AI doing marketing work for us rn

BREAKING: The first AI marketing agent for search visibility is here. It shows where your brand is missing across ChatGPT, Gemini, and Claude... and helps fix it.

@Amrit_Mirch @pmarca AI is leverage, not strategy. Marketing judgment still decides what scales.

@pmarca dont agree use AI but use ur marketing prowess and creativity and you win

@currentbitsNET @JonBuildsHQ Judging contradictions beats scoring them. Scores create fake precision while the mess stays.

Still trying some setups but currently it goes like this The agents don't score contradictions they judge them. A janitor flags maybe-conflicting pairs and queues them. Then an agent decides: keep both, merge dupes, or invalidate one. Only auto rule is the tiebreaker newest wins. Losers get tombstoned, not deleted.

If you’re building in public with under 1k followers: drop your product. Let’s find the underrated builders.

@currentbitsNET Distilled handoffs stop error buildup. Agents stay reliable for long multi-step marketing runs.

@ericDvoss Handoff stays clean because what's passed along is distilled facts, not raw transcripts, so it doesn't balloon as a run gets longer. Each step pulls only the memories that are relevant, so a 20-step run hands off about as cleanly as a 3-step one.

Seeing people pair Grok 4.5 with Fable as the orchestrator. If you've tried it, how's it going? Genuinely curious, not looking for benchmark screenshots, just real experience.

@xikhar Models are close enough. For agency marketing, pick the cheaper generous one every time.

@ericDvoss The models are more or less equally good on most tasks, and one of them is cheaper and more generous.

Anthropic could never match this

@kalashvasaniya AI answer visibility looks good on paper. Users look good on the bank account.

do you understand what just happened?? a fully bootstrapped team with 5m ARR just launched the AI agent that does what $100k/mo agencies do it can: - audit a brand's entire AI search presence - find every gap across ChatGPT, Gemini and Claude - identify why competitors show up and you don't - write every fix automatically - implement everything without human input - run continuously in the background the future of marketing is here

@assafronen Marketing first is right. AI speeds ad creation but still scales weak messages that bring users who bounce.

Start with marketing. We used to spend days producing static ads with agencies. Now an AI pipeline takes a competitor's ad and turns it into an on-brand one in minutes. One of our best-performing creatives last quarter was built this way — for a rounding error.

I'm the President of the HelloFresh Group and sit on our management board. Over the last 6 months we've quietly rebuilt how we operate with AI. Not a pilot. Not a slide in a board deck. Real P&L impact. Here's what's working, what broke, and what nobody warns you about 🧵

@ZaidAD_ That handwriting test kills most AI openers before they ever hit send.

the AI scraped opener is everywhere now and i dont think people realise how obvious it is. heres the test i run before any campaign goes out. if i was handwriting this email, to this one person, would it sound like this? nobody handwrites "i saw youre building a marketing automation platform for agencies, with AI-driven cold email and automated replies tied into a CRM." you would never say that to someone's face. youd say "saw what youre doing with Synteria, love it" and then get to the point. feeding a scraped homepage into AI and asking for an opener per lead does the opposite of what people think it does. it doesnt make the email feel personal, it makes it feel processed. the prospect ends up reading a paragraph about themselves they didnt need, written by someone who clearly didnt need to know it either. ive run plenty of campaigns with no AI personalization at all and they did fine. it isnt a requirement. it can help when its done properly, but most people arent doing it properly, theyre pasting a homepage into a prompt and mailing thousands of people with the output. if you cant get AI to write it the way youd write it by hand, just dont use AI.

@synthwavedd This should be interesting.

🚨 SCOOP: Use your Fable 5 quota while it lasts. Anthropic are working through the final stages of preparations for the launch of Claude Opus 5, with plans to launch as early as later this week (this week and next week encompass the 2-week target window). As it stands, they are *not* intending to extend Fable's inclusion in subscriptions beyond July 19th. A successor to Fable 5 is planned for August to compete with GPT-6, and for now they think they can get away with launching an Opus model not too far from Fable but cheaper to serve.

@araseb_ Grok and Gemini must feel left out.

Why does every AI start with “C”? ChatGPT, Claude, Codex, Copilot, Cursor… coincidence ??

@kaaaash____ No. Real building still needs a full OS and proper tooling.

Can an iPad really replace a MacBook?

@DMSCoding11 Linux. Full control beats polished friction every time.

Developers, which OS are you focusing on right now?