OpenAI and Anthropic are great examples of freemium vs full-paid models OpenAI → freemium 800M WAU, $20B ARR Anthropic → full-paid 20M WAU, $15B ARR If you choose freemium, you need to serve 40x more users to grow profit by ~30%

The choice of messenger by OpenClaw / Manus says a lot Telegram is getting a massive inflow of new users thanks to its Bot API X gets nothing If X doesn't ship a Bot API soon, it's going to miss the AI-agent messenger market entirely

90% of founders come to X to build and stay to farm likes

Startup idea: 4o wrapper (they removed it from the web, but not from the api) You'd get millions of users missing that model

X's algorithm got insanely sensitive I liked 3 memes and now that's all I see lol

Everyone talks about openclaw like it's fully local, forgetting that all your data still goes to the model's api

Why is X full of posts saying it's the worst social network?

Technical interviews are basically dead Latest models solve ~3500-level algorithmic problems, while the average leetcode problem is around ~2000 Interesting to see what employers do next

Gemini 3.1 Pro is way more powerful than Opus 4.6 Can't wait to try it in antigravity

Just shipped gdpr-friendly cookieless analytics for my blog builder 12 hours fighting with ClickHouse 🫡

AI models launched in 2026: • Claude Opus 4.6 • Claude Sonnet 4.6 • GPT-5.3 Codex • Gemini 3 Deep Think • Gemini 3.1 Pro • Qwen 3.5 • MiniMax M2.5 • GLM-5 • Grok Imagine 1.0 • Grok 4.20 • Seedance 2 We're not even done with Q1 This year is going to be insane

The openclaw hype is fading, and we still haven't seen a single real use case. No one actually wanted to use it. The real goal was to post a wow screenshot and farm impressions and likes.

Never try to sell anything to indie hackers They'll burn a weekend rebuilding it instead of paying $10

Interesting trend in AI Big companies are shifting from general models to specialized ones (Codex, Claude Security, etc) Building a big general model is cool Solving one specific problem really well is better

I struggled with adding a card requirement for my $0 trial I was afraid it would kill conversion Turns out, without it ~90% of users churn immediately without ever trying the product It's not about money It's about intent

Fans when their AI company drops a new model: - "it crushes all the benchmarks!" Fans when any other company drops a new model: - "benchmarks don't mean anything"

Adding revenue sharing on X was the worst decision.

That's why building SaaS for indie hackers is mostly pointless You're inside an AI bubble where anyone can vibecode your product for free But once you build for non-technical users, that question disappears No one asks "why should I buy this if I can vibecode it?" Most of them

What an amazing coincidence that 90% of people praising OpenClaw are also selling a course about it.