Notes from the packing floor
The Blog.
Models, automation economics, and the occasional strong opinion. Written by humans who ship this stuff, with AI pushed to work on the research.
“Fable 5, Free for X More Days.” The Deadline That Never Dies
Anthropic has extended free Fable 5 access on paid plans at least twice in five weeks. Part supply caution, part the smartest marketing loop in AI, and I think it's a bit of both.
The July 2026 Frontier Scorecard: Fable 5 vs GPT-5.6 Sol vs Kimi K3
Three frontier models launched or dominated within five weeks. One table, the honest trade-offs, and which one belongs in your stack.
Kimi K3 Is the Most Important Release of July, and It's Not Close
Moonshot's 2.8-trillion-parameter open-weight monster is beating GPT-5.6 Sol on agentic benchmarks at half the price. The market just changed.
More With Less: How Chinese Labs Engineer Around the Compute Gap
Under U.S. export controls, Chinese labs can't buy their way to the frontier, so they engineer their way there. Sparsity, attention tricks, and aggressive open-weight pricing are rewriting the rules.
Freelancers Are Using AI to Find AI Automation Gigs. We've Gone Full Ouroboros.
A popular n8n template scrapes four job boards and auto-writes proposals for automation work. The snake is eating its own tail, and I respect the craftsmanship.
The 7 Automations Small Businesses Are Actually Paying For in 2026
We went through the live freelance market: Upwork categories, Fiverr gigs, n8n templates. Seven jobs, and that's where the money actually is.
OpenAI's Quiet Play: ChatGPT Work, Codex, and the Ona Acquisition
Everyone watched the model launches. Meanwhile OpenAI quietly bolted together an agent stack aimed straight at the office.
Zapier vs Make vs n8n in 2026: Speed, Value, Control
The rule of thumb the whole industry quietly agreed on, and how to work out which side of it your business sits on.
xAI Is Now SpaceXAI, and I Have Feelings About It
A rocket company rebranded its AI lab and joined the Nasdaq-100 the next morning. That running order matters more than the logo.
Google DeepMind Got a “C” in Safety. The Letter Grade Is the Least Useful Part.
The Future of Life Institute handed frontier labs report cards. The grade is a headline. What sits underneath it is the bit worth reading.
Why AI Receptionists Became the Hottest Automation Gig of the Year
Voice agents answering business phones went from demo to default in about 18 months. The economics explain the whole thing.
How Automation Work Gets Priced, and How Not to Get Burned Buying It
Per-project, retainer, or per-execution pass-through: what each pricing model quietly tells you about the builder, from someone who sells this stuff.
“Mythos-Class” AI, Explained Without the Hype
Anthropic's Fable 5 launched a new capability tier above Opus. What actually changed, what it costs, and whether it's worth it for your business.