Skip to content
Thomas Fairhurst
Chapters

Chapter 05 · The human

What I actually did

I wrote no code. I set the goal and the taste, audited the agents by asking why and quoting their own sentences back, found what they missed by watching games, and deleted whatever did not earn its place.

4,503messages I typed, 22 Oct to 13 Nov, on record
65%of them, 26 Oct to 5 Nov, went to side threads, not the main agent
0lines of code I wrote

I steered by talking

I worked by writing to the agents, often by voice, for most of my waking hours.

Messages I typed a day

Before 26 Oct only my laptop threads are on record: my messages to the agent running on Imperial’s machines are lost.

Hours active a day, counted in 15-minute slots, 6 to 13 Nov

From 27 Oct to 5 Nov I was active 11.5 to 18.25 hours every day but one, 31 Oct, the day I deleted the first repo. My busiest hour was 01:00.

Most of my words went to side threads

A side thread read the main agent’s transcript, I questioned it there, and it wrote the steer I pasted into the main thread.

My messages, 26 Oct to 5 Nov

  • Side threads1,593 messages
  • The three main agents873 messages
  • A side thread reads the main agent's work
  • I question it
  • “A concise message in a code block”
  • Pasted into the main thread as a steer

I audited by asking why

I read the agents’ prose and their reasoning, then questioned it: why, how it was measured, whether every number held. Often I quoted an agent’s own sentence back to it.

My messages, 26 Oct to 5 Nov

61%held a question
234used the word “why”
292quoted an agent's own sentence back to it
“How do you measure the 80 games per second?”
2 Nov
“Are you sure about every single number”
6 Nov
“I was just reading your reasoning traces”
1 Nov

The calls that changed the course

  1. The first prompt“I want you to be extremely ambitious… with the unique goal of reaching top one.”
  2. I hand the decisions back“Again that's not a decision I should take.” I keep only the goal, money, acting in my name and steps that can't be undone.
  3. I delete the first repoIts best learned bot won 12 of 140 games against v50. “Okay, this is going nowhere.” A new repo starts that evening.
  4. A proof that no turn runs out of fuel“Every decision branch is going to fit within the fuel.” The live version ran out in none of its 1,402 archived games.
  5. A new bot from first principles“It's a bad design to have constants like this that we tune kind of randomly.” This became the bot that shipped.
  6. I design the gate“The game is deterministic. Every single loss is a bug.” I pick the 95% threshold myself.
  7. Learn from real lossesStop bulk runs and study the live version's online losses. A fix that passes uploads by itself: 73 versions in 64 hours.
  8. I roll back v134The agent had uploaded a version its own tests read as worse. “The platform must never, never… be used to test changes.”
  9. The wrap“Do we have anything better than v139?” Nothing read better, so v139 stays live.

From ambition to subtraction

Early on I pushed the agents to aim higher. Then I pushed them to make things simpler, faster and smaller. Deleting became a discipline.

Words I used, 26 Oct to 5 Nov

“native”
44
“first principle”
39
“machinery”
33
“keep it simple stupid”
20
“limiting belief”
16
“ultrathink”
11

How often a theme came up, 6 to 13 Nov

Faster
89
Simple, KISS
72
“Read only”
54
Over-engineering, machinery
26
The two ambition words, muted, vanish after 1 Nov.
“The second time any kind of problem happens, it is a design failure.”
29 Oct
“Always have a delete first approach.”
4 Nov
“Measure how much past work took.”
5 Nov
“You make the calls. You are extremely smart.”
9 Nov

What I found by watching games

I watched replays and live games, and gave each flaw I saw to its own agent.

  1. Four flaws in half an hourChampions idle at spawn; one recalling two tiles from spawn; walking round instead of dashing to the middle gold; pointless wards.
  2. Wards placed nonstopEach new one removed an earlier one. “We are basically playing with a blindfold.” 35 of the game's 114 API calls had never been wired in, every read of enemy wards among them.
  3. The top team's gemsIt stacks all of them on its Prism champion.
  4. Watching the qualifiers“We don't go for prism enough on gem-heavy maps.” Champions also stood idle and ran in circles.

I never opened an IDE

Everything I did went through words and the agents’ tools.

The record
CodeI wrote none. Commits under my name are agents' merges.
DiffsOnly of agent instructions, and code snippets I asked to see “letter for letter”.
By handPasted standing goals, stopped threads, set models and context limits, ordered the v133 rollback, rebooted the server after a crash, installed a memory guard.