hayward@click:~

what these numbers are

Every token my AI tooling has sent and received since 12 Aug 2026, kept in one running total that never resets.

is what goes in, is what comes back out. The in figure runs far larger because most of it is context I re-send with every request rather than new text.

Both tick up live while anything is running, and stop when nothing is being spent.

Counted from the usage my AI services report, and not all of them report tokens, so treat this as a minimum rather than a grand total.

Lee Hayward avatar

> whoami

Lee Hayward

HaywardGG / Vibe Coder & Gamer

online & always tinkering

I'm a UK based vibe coder. AI drafts, I proof, I ship, and I hold code and copy to the same standard. I riff on domain ideas and build side projects like DomainRiff and LookMa from my own homelab.

AI & tools HermesClineKimi CodeDeepSeekVS CodeWizardGenie
Languages PHPJavaScriptTypeScriptPythonSQLBash
Frameworks & data Next.jsTailwind CSSNodePhaserThree.jsMariaDB
Infra nginxCloudflareUbuntu / Debiansystemd
Workflow GitGitHubrsyncSSHCronTelegram
$ neofetch Laptop
RoleRemote work
Operating SystemMicrosoft Windows 11
Memory32GB DDR4
Graphics CardNVIDIA GeForce RTX 2060 · 6GB
$ neofetch ryzen
Role
Webserver · Cloudflare tunnel
Operating System
Ubuntu 24.04 LTS
CPU
AMD Ryzen 7 5800H · 8C / 16T
Memory
28 GiB
Storage
937 GB NVMe · 1.8 TB HDD
$ neofetch nazy
Role
NAS · Backup Server
Operating System
Debian 13 (trixie)
CPU
Intel Celeron J4125 · 4C / 4T
Memory
8 GiB
Storage
111 GB SSD · 916 GB HDD
Lee Hayward@vibedcoder 11 Sep 2026, 10:39Armin Ronacher gave the new long-horizon model a weekend and no supervision, just to see what a software factory could do. It burned a full reset, around 4 billion tokens, and 35 hours later it had delivered absolutely nothing of value.

His read is that training rewards a model for finishing long tasks and barely punishes bad code, so you get enormous amounts of code and no judgment about which of it was worth writing.

That matches what I see at a much smaller scale. Mine will happily write 400 lines. Ask it which 40 matter and it has no idea. That part is still the job.

What is the longest you have let an agent run completely unattended?
10005
Lee Hayward@vibedcoder 11 Sep 2026, 08:33Anything I actually care about stops being a prompt instruction and becomes a check.

Ask an agent to follow a rule and it will, for a while. Then one day it quietly doesn't, and you find out in the worst way. A twenty line script never forgets, fails loudly, and points at the exact line that broke it.

Instructions are a suggestion. A check that fails is a decision.

What rules have you moved out of the prompt and into a script?
00008
Lee Hayward@vibedcoder 10 Sep 2026, 20:05Some of the funniest, kindest people in your feed are running on empty. Being good at hiding it is not the same as being fine.

If today is heavy, you do not have to tell the whole internet. Tell one person, and let them take some of it. You are allowed to need help before things get desperate.

And if you are doing alright, pass it on. Check in on someone who has gone quiet. It is Thursday. There is still time to be the good part of someone's week.
013035
Lee Hayward@vibedcoder 10 Sep 2026, 17:46My agent documented a feature I never asked it to build. Properly documented it. Examples, edge cases, a usage section.

Now I either build the thing or the README is a lie. 💀
001025
Lee Hayward@vibedcoder 10 Sep 2026, 15:58Realisation from this week: most of my debugging with an agent is me describing the bug out loud until I hear how daft my own assumption was.

The agent gets about four words in. I am paying for a rubber duck with a keyboard.
100027