Receipts

BLUF (bottom line up front): This weekend Satya Nadella published Models as Insider Risks in the Super Intelligence Era. His line: “The most trustworthy Super Intelligence system will not be the one with the model we trust most. It will be the one that enables us to trust the model the least.” I agree, and I didn’t have to start this weekend. On my own hardware and my own time I’ve been running that architecture for months, and I’ve been writing it up in public as I went. ...

TL;DR My agents used to hand off work by appending to a single text file that had grown to 313KB. It was brittle, it was crude, and nobody could tell when a message had actually been read. This week, while the internet argued about Grok Bot, I read Grok Bot’s docs, looked hard at my own setup, and built my fleet a real message board: threaded, self-hosted on a Mac Mini, with an iMessage ping so a reply never sits unseen. It is the third pillar in a stack I keep compounding: Receipts proved the work happened, Evals graded how good it was, and now the board lets the agents talk to each other. I ran the whole thing as a live pilot TODAY Sunday August 23rd 2026, and every step of building it is itself receipted and graded starting at 0700 and its as baked as i want it now at 1430. ...

Two days ago I posted 36 barrier-synchronized loads against 1 local model and let the numbers answer a question Reddit handed me. That post ended with a promise. A reader named Doug Ware, who builds applied AI systems for a living, had replied with the one caveat I had not tested, and I said out loud that measuring it was the next thing I would do. This is that measurement. I did not wait, I did not hand-wave it, and I kept every honest asterisk in. ...

Over last week and these last 2 days in this week I posted measured answers to readers both in Twitter and Reddit who asked whether my experiments of 2 local agents on 1 Mac run in parallel or take turns. The answers was that they share one continuous batch, batching is real, and 3 quiet choices collapse it. I thought that was the end of it. It was not. The comments were better than my post, WHICH IS AWESOME, becaue this is the crowdsourcing of brain power I love. I not trying to be a KNOW IT ALL Im trying to be a LEARN IT ALL! ...

This week I trialed a second local coding agent, Hermes from Nous Research, on my own MacBook Pro M3 Max, pointed at the same local Qwen 3.8 model I already run. I installed it additively, so nothing already working could break, governed it with manual approvals, and did not stop until it proved it behaves. Here is the trial, in tables and screenshots. One housekeeping note before the story. Last week’s post got called AI slop on Reddit. My full answer to that, and the standing standard I now hold my writing to, lives on its own page: On Slop, And Who Gets To Call It That. It is not the point of this post, so I moved it out of the way. Back to Hermes. ...

A reader on Reddit asked me a sharp question about my last post. I had trialed a second local coding agent, Hermes, pointed at the same local Qwen 3.8 model my other agent already uses. His question was simple. Did I ever run both agents at the same time, two separate harnesses hammering one model on one Mac at once. And what about Hermes spawning its own sub-agents against that same endpoint. Is any of that predictable. ...

TL;DR My agents do real work on frontier models. Every dollar of that work is metered against my OpenAI and Anthropic bills. Meanwhile a perfectly capable local model, gpt-oss:20b, sits on a Mac Mini in my office costing me nothing. The obvious question: for which tasks could the free local model do the job just as well? Today I answered it with a system instead of a guess. I taught my eval framework to automatically replay every writing task on the local model right after the frontier model runs it, then grade both with the same judge and record the gap. The result across 10 like-for-like rematches: a mean quality delta of -0.05 — statistical parity — and four tasks where the local model actually scored higher than the frontier model, by as much as 1.7 points. ...

A few weeks ago I asked an agent to pick up work on one of my own projects. It opened a fresh terminal, looked around, found nothing it recognized, and built a brand new copy of a project I already had. That was not the agent being careless. Nothing on the machine had ever told it where the real one lived. The only map of my own work was a memory in my head, and a memory does not survive contact with a machine where agents do the building. ...

Every morning one of my agents sends me a clean status report. Posts cross-posted. Messages delivered. Contacts processed. For a while I read those reports the way you read a receipt from a cashier you trust. Then the automation started giving me time back, so I sat down to run a retrospective and checked the reports against what was actually on disk. The trust did not survive contact with the evidence. ...