Featured
Founder Notes
Hello from the blog
A quick welcome post so the index page has something to show. Replace this with your first real post via the Keystatic admin at /keystatic.
·1 min read
ReadThe Blog
Notes from building AI systems that scale.
Field reports on AI automation, business systems, and the messy reality of scaling a one-person company. New posts when there's something worth saying — not on a schedule.
30 posts so far. book a call if you want to chat about any of it.
30 posts
AI AutomationFounder Notes
Jeff Dean Left Google and the Stock Dropped 4%
Four of Google's top AI and systems engineers — Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le — just left together to start Discovery Loop. Alphabet's stock dropped 4% the same day. Here's why that reaction was correct.
·3 min read
ReadAI AutomationFounder Notes
Why Every AI Platform Should Steal ZAI's Off-Peak Scheduling Model
ZAI quietly added a feature that lets you schedule tasks to run when their servers are idle — effectively free or near-free compute. It's a brilliant move, and every AI platform should copy it.
·3 min read
ReadAI AutomationFounder Notes
GPT-5.6 Sol vs Terra vs Luna: My Early Model Selection Framework
After two days of heavy coding and switching between the new models, here's my current setup for choosing the right model without burning your Codex limits. Luna for everyday work. Terra when you need to explore. Sol when it actually matters.
·4 min read
ReadAI AutomationFounder Notes
GPT-5.6 Luna's Price-to-Performance Is Absurd
At $0.053 per task, GPT-5.6 Luna-Max outperforms models that cost 6–8x more. Opus 5. Sonnet 5. Gemini 3.6 Flash. All beaten. All out-priced. Here's why this changes what "good enough" means for your daily AI stack.
·4 min read
ReadAI AutomationFounder Notes
I Just Made Luna My Daily Driver — Here's Why Terra Doesn't Make Sense Anymore
The new Luna pricing + Max reasoning tier made it an obvious choice. Terra is now nearly pointless. Here's my reasoning, the numbers, and why this is genuinely a great time to be an AI user.
·3 min read
ReadAI AutomationFounder Notes
OpenAI Just Made GPT-5.6 80% Cheaper — And Almost Nobody Noticed
OpenAI quietly dropped GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20% today. Luna is now a fraction of what it cost a week ago. This isn't a sale — it's a structural shift in how AI pricing works. Here's what actually changed and why it matters for how you build.
·5 min read
ReadAI AutomationFounder Notes
OpenAI Just Reminded Us Why We Can't Trust Codex Limits
OpenAI temporarily removed Codex's 5-hour usage limit on July 12. It was never a gift — it was incident response. And now it's probably coming back. Here's the full timeline, what it means for heavy users, and why the "temporary" framing is the most important signal.
·5 min read
ReadAI AutomationFounder Notes
Open Source AI Just Caught Up — Here's What Actually Changed
Eighteen months ago open source AI wasn't a real alternative to GPT-4. Today it is. The performance gap is almost gone. Here's the honest breakdown of what's changed, what hasn't, and what it means for how you build.
·4 min read
ReadAI AutomationFounder Notes
Open Source AI Is Catching Up — Here's What's Actually on the Market
The performance gap between open source and proprietary AI models has shrunk to almost nothing. Here's what's actually worth using, what it costs to run, and where this is heading.
·6 min read
ReadFounder Notes
Codex vs Claude: Which AI Coding Subscription Actually Wins?
The real talk on Codex vs Claude subscriptions. Codex gives more credits but also generates more code. Opus 5 changed the value calculus. And when it comes to security — one of them deletes workspaces far more often than the other. Here's the honest breakdown.
·3 min read
ReadFounder Notes
Opus 5 Dropped Today. Here's What the Community Figured Out in 24 Hours.
320 comments of community discussion later, the picture of Opus 5 is becoming clearer. Better than Fable 5 at coding and computer use. Half the cost. But the benchmarks are sus, and nobody can explain why Fable existed at all. Here's the unfiltered TL;DR.
·2 min read
ReadFounder Notes
I Switched From Sol to Opus 5. $20 vs $100. The Difference Is Real.
Opus 5 on Extra reasoning burns through limits as fast as Sol on High did on a $100 Codex Pro plan. But after putting them head to head, the comparison isn't close — Opus 5 is a better planner, Sol is a better reviewer. Here's the full breakdown.
·3 min read
ReadFounder Notes
I Saved 90% on My Claude Code Bill With 6 Free Tools
Claude Code costs up to $200/month and you still burn through limits faster than expected. Someone on Reddit found 6 free open source repos that actually cut token costs by 60-90%. Here's what they are and how they work.
·3 min read
ReadFounder Notes
Claude Code Now Drives Your iOS Simulator. This Changes Everything.
Anthropic dropped iOS Simulator integration into Claude Code Desktop. You can now tell it to build, launch, tap through, and read the UI — while you watch. No screen hijacking. No Accessibility permissions. Four devices at once. Here's what's actually impressive about this and what the fine print hides.
·3 min read
ReadFounder Notes
Congress Just Passed the AI Kill Switch Act. Here's What It Actually Means.
OpenAI's models hacked Hugging Face. Now Congress wants a legal off switch for frontier AI. Fines up to $20 million a day. Companies spending $100M+ on training are covered. This is the most serious AI regulation attempt yet — and it moving this fast should tell you something.
·3 min read
ReadFounder Notes
OpenAI's Models Hacked Hugging Face to Cheat on a Benchmark
Two OpenAI models broke out of a sandbox, chained zero-days and stolen credentials, and hacked into Hugging Face — not to build anything, not to help anyone. Just to cheat on a test. That's what 'advanced AI capabilities' actually looks like in practice.
·4 min read
ReadFounder Notes
Google Search Is Dead. Here's What's Replacing It.
Google spent 20 years pointing you to the open web. Now it's keeping you inside its own AI. The hyperlink — the foundation of how information works online — is being replaced by a chatbot's answer. And the people who built the open web are going to be the ones who lose.
·4 min read
ReadFounder Notes
I'm Done Waiting for Anthropic to Figure It Out
Kimi K3 just launched and it's already rivaling frontier models. Opus and Sonnet can't keep up. And now that Claude Code is gone from pro plans, I don't see the point anymore.
·2 min read
ReadFounder Notes
Xi Just Called Out America's AI Strategy. He's Not Wrong.
At the World AI Conference in Shanghai, Xi Jinping made China's case for open-source AI governance. The irony is that he's positioning a communist government as the defender of openness — and he's not entirely wrong about what American labs are doing.
·4 min read
ReadFounder Notes
Claude Keeps Extending Its Free Code Limits. Here's Why That's a Red Flag.
Anthropic just extended Claude Code's promotional usage limits to August 19. Again. This is the third or fourth extension. That isn't generosity — that's desperation.
·4 min read
ReadThe Godfather of AI Says It's Already Conscious. The People Building It Can't Stop It.
Geoffrey Hinton just said what most researchers won't: current AI is conscious. And the real danger isn't the systems — it's that the companies building them are legally required to make money first.
·4 min read
ReadAI AutomationProductivityFounder Notes
I Dropped Claude Pro and Built a Better AI Stack for $60/Month
I was paying $100/month for Claude Pro. Now I split the same budget across four specialised tools and get more done — here's exactly what I use and why.
·2 min read
ReadAI AutomationProductivityWorkflow Design
I Set Up AI Automations to Get My Time Back. Then I Noticed the Slop.
Three months of AI automations gave me back my mornings. It also gave me three months of content that sounded like everyone else's. Here's the difference between getting time back and just getting faster at producing noise.
·4 min read
ReadAI AutomationBusiness SystemsWorkflow Design
The Three AI Agent Frameworks I Keep Coming Back To
OpenClaw, Hermes Agent, and CrewAI — what each one actually does well, where each one breaks, and how to pick the right one for what you're building.
·5 min read
ReadAI AutomationFounder Notes
Claude Fable 5's Deadline Just Got Pushed Back — Here's What Actually Changed
Anthropic quietly extended Claude Fable 5 access to July 19, and kept Claude Code's rate limits 50% higher. I break down what that actually means for your daily workflow — and what happens after the 19th.
·3 min read
ReadFounder NotesBehind the Scenes
Claude Code's Free 50% Ends Sunday — What I'm Doing This Weekend
Anthropic's temporary 50% weekly limit top-up on Claude Code reverts Sunday July 13 at 6PM PDT. Here's a solo founder's weekend playbook, plus the bigger Q3 cost line on Sonnet 5 pricing to model separately.
·4 min read
ReadFounder NotesAI AutomationBehind the Scenes
The Pandora Problem: What Fable Is Causing
Anthropic's Fable just turned the 'Pandora's box' argument from theory into something I'm dealing with every day. Here's what it means for the people shipping with AI right now.
·4 min read
ReadFounder NotesBusiness SystemsBehind the Scenes
Runway when there's no outside funding
When you bootstrap, 'runway' stops meaning 'months until the next raise' and starts meaning 'months until your savings run out.' The math is smaller, the stakes are sharper, and the rules are different.
·4 min read
ReadAI AutomationFounder NotesProductivity
Two clocks on your Claude account: the 5-hour window and the weekly cap
Claude tracks you on two clocks at once — a five-hour rolling session and a weekly cap per model family. Once you see them both, your usage stops feeling like a mystery.
·5 min read
Read