Johnny

founding engineer @intelligenceco 🌻 prev. eng in residence @terraincap

[Johnny @johnnyheo]

brood war skills definitely coming in handy these days

effective APM is everything

[Anicet @AniC_dev]

gosh I'm only hiring ex RTS players now

it's literally impossible to do the job I've been doing the last 10 months, and increasingly so if you can't

> parallelize >5 things
> operate at 100 Actions Per Minute

[Johnny @johnnyheo]

@gakonst Way too many changes to keep up with every time I open it. Works alright but I don’t feel very in control of the harness which is the main reason I go for Pi

[Johnny @johnnyheo]

the garden grows 🌻

[Sagar @sagarck04]

update:

i'm joining @intelligenceco as an applied ai researcher!

we're on a mission to build the infrastructure for the one-person billion-dollar company.

agents are doing crazy things, but we're only scratching the surface. @intelligenceco is pushing the frontier of what real agent-led, long-horizon autonomous work looks like.

excited to work alongside @spencersunoo and join the research team :)

shoutout to @ndrewpignanelli, @KhareAbhishyant, and @katmcguir3 for moving fast and making this an easy decision.

lfg!

[Johnny @johnnyheo]

@mitch_troy @thsottiaux @jxnlco That’s just the default you can change it manually via codex -c model_auto_compact_token_limit=<370k if you want>

[Johnny @johnnyheo]

@ApplyWiseAi pretty immediately - the bro skill i shared in another reply helps too

[Johnny @johnnyheo]

Fable and Sol tend to use a lot of jargon that's hard to understand on a first read, which slows me down a lot

i've found incredible success adding the following to agents or claude.md

Write user-facing explanations in clear, concise language without reducing technical precision. Prefer concrete wording over unexplained jargon. Use established domain terminology when it is the most precise choice, and briefly define it when the intended audience may not know it. Preserve material evidence, constraints, tradeoffs, caveats, and uncertainty. Do not rewrite code, identifiers, commands, quoted text, or prescribed formats merely to satisfy this style rule.

and adding personality = "none" to ~/.codex/config.toml

[Johnny @johnnyheo]

@stevencheng it helps me make sense of them, but ymmv!

[Johnny @johnnyheo]

@jppltech works great in codex!

[Johnny @johnnyheo]

@UrMoonDoo claude is a certified yapper

[Johnny @johnnyheo]

@bakardev exactly! right now i'm focused on doing everything i can to keep myself truly informed while moving fast

this practice helps reduce cognitive load and makes context switching between threads much more fluid

[Johnny @johnnyheo]

@aaronvi i'm being a little conservative with my wording in agents.md after seeing people nuke their hard drives

sol's a little too eager to follow instructions, might end up writing worse code

[Johnny @johnnyheo]

Grok 5 will probably take everyone by surprise

[Sebastian Raschka @rasbt]

Upon request, here's an updated version with Grok 4.5 and Meta's Muse Spark 1.1.

Grok 4.5 seems to sit at the Pareto frontier. Good bang for the buck.

(Also added harness info).

[Johnny @johnnyheo]

@notyourdoctor it's in my global

if you're concerned about potential impacts, this skill will also do the trick! https://t.co/bA57MvJMYu

[Johnny @johnnyheo]

@emollick Wtf

[Johnny @johnnyheo]

@elliotarledge Increasingly impressed by Grok 4.5

xAI really did their thing!

[Johnny @johnnyheo]

@rudrank wonder why progress on this model is so slow

[Johnny @johnnyheo]

@OpenAI Let's gooooo

[Johnny @johnnyheo]

@elliotarledge @taziksh grok 4.5 is that good??

[Johnny @johnnyheo]

@ella_schlags Agree! https://t.co/KDvHnqPk1F

[Johnny @johnnyheo]

If you’re going fail at something, fail spectacularly, in the most embarrassing way possible

You learn nothing by playing it safe

[Johnny @johnnyheo]

@catboosted this is good