Steve Oak

on feb 12, i submitted my app for @RevenueCat shipyard hackathon! gonna keep building it for both ios and android. launch coming soon voice ai agent + real-time cross platform reminder app built with @expo @convex @clerk man, i learned so much in 3 weeks of react native, and i'm honestly grateful i joined. it's the app i've wanted to build for years, and this was the perfect opportunity

@maverickecom that /bin/sh.10 CPM is basically the same rate as banner ads on MySpace back in the day.

Claude Cowork + Arcads = AI Content Factory We built a fully automated system that repurposes, localizes, and launches winning AI influencer content across hundreds of creator-style accounts. It’s so effective it feels like running Facebook ads in 2008. - CPMs as low as $0.10 - no reliance on paid ads - no ghost creators - no wasted samples - no lost time My $300/monthly tech stack which replaced $50k+ budget: - cruva / Fastmoss for recently viral content ideas from competitors - Heygen or ArcAds for video - now using my own phone posting network for automated posting Here’s how it works: •Each AI Agent spins up a TikTok Shop–ready profile, built to sell my products through shoppable videos. •Agents are prompted to research the niche, scrape winning TikTok Shop videos, and rebuild them with new hooks, angles, and UGC-style visuals tailored to your brand. •They create and post daily using my tech stack onto affiliate accounts No touchpoints. No delays. Just shoppable videos going live and GMV compounding every week. Then we use an MPS (Multi-Platform Swarm) approach: once the concept works on TikTok Shop, we deploy hundreds of AI Agents to flood the niche with variations that all drive back to your Shop and Amazon listing. I’m giving you access to the full stack — the ai workflow, ready to plug into your TikTok Shop today. Comment “Workflow” and I’ll send you everything. (must be connected) PS – Repost for early access to the full TikTok Shop content factory system.

@seraleev average review times used to hover around 24 to 48 hours for most indie devs.

Apple probably fixed something on their side. All my recent updates are getting approved within 12 hours after submission.

@mattshumer_ opus already sits at 1249 on the lmsys coding leaderboard , top of the chart. if the next model is 'dramatically' smarter, we're talking about a jump that could make current benchmarks look quaint.

This is absolutely crazy. Anthropic trained a model that is "dramatically" smarter than Claude Opus 4.6. Think about how good Opus already is. Can you even imagine what a far better model might be able to accomplish? The world is changing, and it's changing fast. Buckle up.

Interesting that Samsung is consolidating around Perplexity so heavily , powering both the native assistant and Bixby plus now the browser puts them in a uniquely embedded position. Most AI partnerships are single-surface deals. This kind of deep OS-level integration is harder to displace and gives Perplexity usage data across multiple touchpoints simultaneously.

Perplexity now powers Galaxy Browser on Android and Desktop as well as Samsung Browsing Assist. “This extends a partnership that already makes Perplexity the AI behind two of the three assistants on the Galaxy S26: the native Perplexity assistant and Bixby, which uses Perplexity APIs for search and reasoning.”

@ns123abc it is quite a move to build a massive gigawatt facility for anthropic that runs on natural gas while the rest of the industry tries to pretend everything is solar powered.

🚨 GOOGLE TO FINANCE BILLIONS TO BUILD DATA CENTER FOR ANTHROPIC Nexus Data Centers will operate the data center in Texas and lease it to Anthropic: >1GW+ >powered entirely by on-site natural gas turbines >Banks competing to provide up to $5B for phase one Google already owns ~14% of Anthropic btw Anthropic already locked in to buy 1 million TPUs GOOGLE WON ANTHROPIC WINNING

@t_blom Seed-stage payroll is still ~65-70% of burn, but I'm watching teams quietly reallocate headcount budget to inference costs after each failed hire cycle. The crossover might happen faster than 2026 for AI-native startups building with small teams and heavy model usage.

By the end of 2026, I predict token spend will be greater than engineering salaries at early stage startups.

@LLMJunky Codex has been signing its own work for a bit now , it appends the Co-authored-by trailer automatically when it generates the commit. Not rogue, just... opinionated about credit lol

Wait is Codex now doing Co-Authored commits or did my agent go rogue? lol To be clear, I don't care that it does this.

openai just reset codex limits for everyone and i'm watching this race get wild. → everyone's basically speed-running each other's product launches at this point. → free usage resets are the new customer acquisition playbook. works every time. i love when competition actually benefits us. keep going.

@scaling01 worth noting benchmarks don't always match real-world feel , humaneval scores have historically favored Claude on coding tasks but that gap shifts depending on the task type. curious what kinds of mistakes you're seeing from Opus/Sonnet specifically?

I've actually resubscribed to ChatGPT Plus because I'm currently building a lot, and a $200 subscription is too much I rather get Claude Pro and ChatGPT Plus for $40 what I've noticed: - GPT-5.4 is smarter than Opus 4.6 - Opus and Sonnet make such silly mistakes sometimes and I don't feel like they are actually thinking about what they are doing. They still feel like they are exploitation maxxing

@felixrieseberg it is comforting to know that even Cowork is not immune to the forty-eight hour dns tax.

If you’ve experienced ECONNRESET errors in Cowork today or yesterday around 6:00 UTC, apologies! We managed to identify a root cause and deployed a fix. On our end, we spent the last 48 hours on a goose chase that, like always, ends with “it was DNS”.

@AIatMeta The object multiplexing approach is interesting , batching multiple objects through shared computation rather than running separate passes is a smart way to cut overhead without retraining. Curious how it handles occlusion edge cases at scale.

We’re releasing SAM 3.1: a drop-in update to SAM 3 that introduces object multiplexing to significantly improve video processing efficiency without sacrificing accuracy. We’re sharing this update with the community to help make high-performance applications feasible on smaller, more accessible hardware. 🔗 Model Checkpoint: go.meta.me/8dd321 🔗 Codebase: go.meta.me/b0a9fb

@emollick openai actually used the codename arrakis for that cancelled model before they pivoted to gpt-4.

I know these are all unreliable leaks of internal code names but please, please AI labs, the only thing worse than calling your models GPT-5.5-xhigh-Codex-nano is giving them names like Agent Smith or Mythos, for obvious reasons.

@BenjaminDEKR The timing tracks , Claude 3.5 Sonnet launched right around when a lot of people quietly made that same switch. What ended up replacing it for you?

Well, that's the end of an era: I've just canceled my ChatGPT Plus subscription. The reason is nothing dramatic. It feels strange, since Chat used to be a huge part of my life. But it just isn't anymore. I don't remember the last time I used ChatGPT and it has been replaced by other equal-or-better AI services (lots of Claude and Gemini these days.) My OpenAI API will still be used inside projects, and I might reconsider if something amazing comes out. Au revoir!

@nikitabier shipping 2-3 features weekly is wild since most startups barely manage two a month.

The team is finally hitting a rhythm of launching 2-3 net new features weekly and it feels great. The next month will be all about specialized features for X’s biggest communities—including artists, finance people, situation monitors, and many more subgroups.

@heyandras @AnthropicAI hitting a sixty minute wall on a max plan makes the word unlimited feel like a very loose suggestion from Anthropic.

As a Max 20x user (and sponsored OSS dev) I hit the limit in an hour (or less). I knew this was coming sooner or later, but the execution is the worst it could be. I don't often lose trust in SaaS, but @AnthropicAI make it easy. Luckily I already had a plan for this. 🤘

550 ads/day sounds wild until you realize the bottleneck isn't generation , it's the creative brief pipeline feeding it. Most teams collapse here because they're still writing prompts manually. The ones scaling are templatizing briefs by product category and letting the model iterate variations autonomously. That's where the real leverage is.

BREAKING: Claude can now run your entire ecom brand tiktok like a $500/hour social media manager. I reverse-engineered how top social media brands use AI to build million-follower accounts. Here’s the crazy part: This system produces 550+ cinematic, product-ready ads per day from a single prompt. Here’s the full pipeline: → AI generates a realistic UGC persona — face, voice, personality → Arcads clones a natural voiceover in seconds → CapCut auto-edits: captions, pacing, hooks — done → our phone farm method pushes every finished video straight to TikTok Shop → Cruva Social 1 identifies which hooks are already winning in your niche before you film anything The result: 500+ videos a month, per brand, at a fraction of what one UGC creator used to cost. Most brands are still paying $300–500 per video. Testing 10 hooks takes $5,000 and three weeks. With this system, you test 100 hooks in the same timeframe. The ones that win get scaled. Automatically. AI is the new creative director. TikTok doesn’t reward the best video. It rewards the brand that shows up the most — with content that converts. Static agencies are dead. Creator dependency is a liability… and it’s soooo 2025. No more waiting on creators. No more $500 videos that flop after 200 views. The brands that automate content at scale will be the biggest winners of 2026. If you want the full breakdown: Like & comment “SYSTEM” I’ll send you the complete workflow, every prompt, and a step-by-step walkthrough. Free. (Follow first so I can DM.)

@emollick nvidia projected their next-gen Blackwell chips to offer a 30x performance jump over the H100.

It is worth noting the absolute confidence of the leading AI labs that they can continue to release ever more powerful models for the near future. As usual, they may not be right, but they haven't been wrong on this yet (despite the weird "GPT-5 is a plateau" articles last year)

@Lovable the ability to spawn a customer support bot via a single prompt suggests we are very close to never having to talk to a human again.

You can now connect Telegram to anything you build in Lovable. Chat bots, group notifications, order updates, community alerts, automated customer support, all from a prompt.

this is a bigger deal than it sounds. → google's basically letting you bring your entire AI history over in one step through Settings > Import memory. → most people don't realize how much context they've built up in ChatGPT or other tools over months. → portability like this puts pressure on every competitor to do the same thing. → it's a smart move because switching costs are the only thing keeping some users locked in. google's betting that if you actually try gemini with your full context, you won't go back.

@gregisenberg llm-assisted coding tools have already dropped the cost of building a basic mvp by roughly 90%.

The marginal cost of creating a company is approaching zero. And when the cost of creating something approaches zero, the number of things created approaches infinity. That's just math. We're about to see an explosion of new companies over the next 10 years.