ai consulting @every • principal trading • author (Wiley)
Joined February 2017
- Tweets1.1K
- Following137
- Followers7.6K
- Likes5K
reading this and lowkey thinking @danshipper might’ve saved Codex from essentially becoming a VS Code sidebar. apparently they originally built it as a pair programmer, and the night before the gpt-5 launch Dan told the head of Codex "with Claude Code, you just let it rip… we don’t even really look at the code anymore." and now it basically swallowed ChatGPT. wild read.
Codex is one of the most important untold stories in the history of technology.
A year ago, it was a scrappy effort in a quiet corner of @OpenAI. Its cloud product hadn’t found PMF, and the team was starting over from scratch.
Less than 12 months later:
- Codex swallowed the flagship ChatGPT desktop app from the inside (it literally runs as com.openai.codex on your Mac)
- Codex and ChatGPT for Work hit 40M+ weekly users across engineering and knowledge work
- Its leader @thsottiaux was elevated to run all of OpenAI’s core products, including ChatGPT, Codex, and the API
How did they pull it off?
They trained their own dedicated model line, separate from ChatGPT, with researchers and product builders sitting on the same team
They built four different products in a single year (cloud → terminal → IDE sidebar → a desktop app with no code editor at all) as every model leap made a new form factor possible
They used Codex to build Codex—so a handful of people could prove a new product in weeks instead of arguing for it in meetings
I had a front-row seat to the whole arc—from their failed Cloud launch, to OpenAI’s offices the night before GPT-5 dropped, to the day Codex took over ChatGPT.
And I spent months interviewing @gdb, @thsottiaux, @embirico, @mia_glaese, @aambrosino, and @edbayes to get inside how it all went down.
Here is the full, behind-closed-doors story of the team that remade OpenAI:
every.to/p/how-codex-happene…
We’re pushing the boundaries of how stories like this are told.
This piece is completely agent-native—designed to be read alongside your agent inside ChatGPT’s in-app browser:
- Every interview transcript is included in full: Click any quote and you (or your agent) can see the exact conversation it came from.
- Built for your agent to explore: Ask your agent to pull out product lessons for your team, cross-examine the transcripts, or find insights I missed when writing the piece.
It’s a story about the future of software—built to be read with an agent:
every.to/p/how-codex-happene…
this is so damn good
Introducing ChatGPT for Dishwashing
GPT-6 achieves state-of-the-art results on dishwashing benchmarks for long-running jobs and difficult stains.
The Navier–Stokes Millennium Prize Problem is a $1 million challenge about whether the equations describing fluid motion can break down. In this video, I accidentally found a counter-example while trying to optimize my dishwasher. But in reality, OpenAI used 10,000 AI agents to find a proof in 88 hours: openai.com/index/navier-stok…
Parody of the original GPT-6 Astra ad
@OpenAI #ChatGPT_Partner #chatgpt
Credits:
Written and Directed by Joma
Co-Writer - Henry Connor Coan IV @coan_henry
Production by Dauntlus Studios @dauntlus.studios
Producer/Co-Director - Marcus Liew @dauntlus
Camera Operator - Amos Lee @amoselijahlee
Art Director - Alethea Soo @aletheasoo
Art Assistant - Alicia Lim @aliciatkl_
Assistant Editor - GPT-6 Astra
Music: Tensions Run High by Soundridemusic
youtube.com/watch?v=Ly9H63SL…
this is what happens when you vibe code a trading system and never bother to learn how trading works.
you can take a trade with no edge, size it like an idiot and still make money. do that long enough and you’ll convince yourself you’re even good at it.
we’ve all seen traders adding to losers because “it always comes back.” for weeks, maybe it does.
they dig themselves out, finish green and take that as confirmation that adding was the right call.
then they get the trend day that doesn’t come back and blow the entire account.
every recovery before that taught them to keep doing the thing that eventually took them out. the risk was there on all those green days. it just hadn’t shown up yet.
that’s the problem with treating short term profit as the whole objective. the market can keep paying you while you learn the wrong lesson.
even a real edge doesn’t get you out of this.
you can have positive expectancy and still blow up because you’re trading too big to survive a perfectly normal losing streak.
knowing something makes money on average is not enough to know how much of it you should do, or whether your account can handle it.
of course you can train a system on pnl. how much risk did it take to produce that number? over what period? can it survive the losses that haven’t appeared in the sample yet?
a few green trades don’t answer any of that.
putting a hard limit on the damage doesn’t prove the behaviour inside those limits is worth learning.
this project also gives it the same reward whether it makes a $.01 or $10. that’s another problem, but it’s secondary.
you could make the reward perfectly proportional to pnl and still be rewarding it for a stupid risk that happened to pay off.
most traders have had a winning trade they’d have been better off losing, because getting paid convinced them to do it again with more size.
you’ve taken the mistake a trader is supposed to outgrow and made it the training objective.
Years of screen time don’t guarantee years of learning. If your review process is basically “made money, good trade. lost money, bad trade,” you can spend a decade getting really good at the wrong things.
Brent needs no introduction. Highly recommend this one. Just grab a copy.
My new book is out!
TRADE OUTSIDE THE BOX: Advanced Thinking for Professional Traders
Go beyond fundamentals and technicals and use second-order thinking, poker strategy, clinical psychology, AI, LLMs, and a better understanding of ergodicity, narrative contagion, the multiverse and much, much, much more.
If you think like everyone else, you will perform like everyone else.
I have a VPS just for Fable 5
> it checks the API every Planck time
5.39e-44 seconds, the smallest unit of time in physics
> if Fable is offline, nothing happens
> if online, 1,300 agents deploy instantly
> there’s a list of 26,413 questions for Fable
sourced from random YouTube replies, shower thoughts, online IQ test screenshots and 2am notes app entries
> all worded to extract maximum amount of knowledge
> fresh API key backed by a home equity line
bank thinks I’m renovating my kitchen
I am not
> 5-10k calls to API in parallel
> dataset will begin populating with Fable answers
> even if they shut it down again, I will at least have the first snapshot
if you aren’t doing this, you aren’t serious enough
I have a VPS just for Fable 5
> it checks the API every 60s
> if Fable is offline, nothing happens
> if online, the program starts
> there's a list of 400+ questions for Fable
> all worded to extract maximum amount of knowledge
> fresh API key with $500 limit on the VPS
> the moment Fable is online, program starts
> 5-10 calls to API in parallel
> dataset will begin populating with Fable answers
> even if they shut it down again, I will at least have this
if you aren't doing this, you aren't serious enough
Most traders place stops where the thesis breaks… and still get stopped out by normal noise.
You already know to place it at structural invalidation. That part’s fine.
What usually gets missed is whether that level is actually outside normal movement on the timeframe you’re trading.
You can have the structure right and still have your stop sitting inside regular volatility.
ATR is a decent proxy for expected noise.
Once the structural level is defined, I want to know whether the stop sits outside normal movement for that timeframe.
1x to 1.5x ATR buffer around that level is a decent starting point.
Then check where the stop actually lands. If it’s on a round number or right above an obvious H/L where stops cluster you’re asking to get run so you'll want to clear that too.
Your stop has to cover three things:
- structure (thesis invalidation)
- expected noise (volatility)
- liquidity (clustered stops)
Take the furthest of the three.
That’s your real stop distance.
Size your trade around that.
If that distance destroys your sizing or the RR looks shit the trade probably isn’t worth taking.
Same thing with targets. If your stop is built around the normal movement of the timeframe, your target has to be realistic for that same timeframe too.
Way too many stop outs come from treating structural invalidation like it’s the whole story.
It’s just the first piece.
Ryan Wright retweeted
One of those books I wish I’d had alongside Alpha Trader when I was first started trading perps / futures in 2018.
Nearly nine months into this journal, the metric I'm most satisfied with is max drawdown: just under 7.5%, with 60%+ account growth on my perps/futures account - it’s not flashy but steady honest work - me, against myself, earning the right to scale size.
Prior to May, drawdown stood at just 3.9%, before I began building larger swing positions.
Across nearly 200 trades since sept, the focus has remained the same - cut losses quickly, refine execution and prioritise risk management over win rate.
The most effective adjustment has been initiating a trade with a pilot position and only adding size once a trade begins proving itself, this gives me a better feel for the rhythm of a market - I’ve been working extremely hard on narrowing the variance of trade execution and efficiency.
The key lesson has been that profitability is driven less by being right frequently and more by managing losses well, I’m currently hovering at a 35% win rate due to cutting trades early only to step into the same idea again upon increased probability (timing issue)
It’s a continuous work in progress, but encouraging progress nonetheless. More importantly I understand the pathway to achieve higher levels… all in the process.
Looking forward to what Q2(June) and Q3 bring, I suspect there will be some decent volatility ahead.
Ryan Wright retweeted
Review of @baynkr's book — and what I've applied to my own trading
TLDR: Despite your experience in markets and your future goals within it, this book is a must read if you want to improve.
(This won't be a summary, just buy the book)
1. It has made me view trading in a completely different way.
Entry, stop, target - this is how most beginners learn to trade. You see single trade pnl's on the timeline and start to believe it is possible to trade a little on the side as some fun side hustle.
Well that is not gonna happen, look at the title again, you are trying to build a business here.
The first few chapters opened my eyes as to why one should even deserves to make a living from trading; What even is an edge?, Should I be outsmarting HFTs?, Are my current strategies built around my analysis or their necessities?, Who is my counterparty here and do I understand them-which one of us is the sucker?
If you read the first few chapters you'll get a better understanding of the answers to these questions
Another key point is that you will rewire the way you think about your longer term performance. You'll stop thinking about singular outcomes and will focus more on the process.
One of my favorite parts is reading about the fact that actually outsmarting your previous common mistakes is a more rewarding feeling than a single fat pnl, because an improved process makes you the better trader, not the single pnl.
Limit orders is another subject I got completely wrong and now I know why. In the past I always used to be happy if my naked limit order would get instantly filled, now I don't.
Ryan highlighted a key point here so I went back through my journal and I have indeed been buying too many lemons instead of peaches.
Usually the unclear but structure and planned trades are the best ones. If you are still working with confirmations only go and read the book and you'll find out why you should actually act more confident when the information is incomplete. You should operate in the fog.
We humans are actually confirmation seekers, Ryan explains why through a biology lens and how traders should work with this not against it because you should not be playing the blocking human tendencies.
Certainty is a luxury for amateurs
Another thing I did not really understand before is how misleading backtests can actually be. It gives you certain numbers in an uncertain game where the rules change every minute.
Besides that its eye opening to read the difference between traders with a similar edge but their eventual pnl once the friction is subtracted. Its a negative sum game, I already knew that but not that it was this crucial to lower your friction tax -> Again, to make a living as a trader you should be running it as a business!
2. From theory to practice
This book isn't just informational, not just a fun read. It gives you a lot of frameworks the professionals use on a day to day basis.
I have fully reworked a few of my sheets:
- Daily report card
- Single trade plan
- Pre-Execution sheets (pre mortem!)
- Trade closure sheet
- Trade debrief
- System 2 that reviews system 1 (Such a must to make sure you are actually reviewing data worth reviewing, in the right regime)
- Weekly reviews
- If -> Then sheet
If this sounds a lot to you...... didn't I say before that this is a business, with this you are just getting started ;)
The reason you need systems is because you will sabotage yourself without them. You won't be more disciplined next time because you wrote it down after your last loss. You need protocols BUT that system 2 that checks all these pre defined rules is a MUST!
3. My favorite aspects of the book based on my personal tendencies / common mistakes.
- "Contrarianism for its own sake is just first-order thinking in a costume." Man have I suffered a lot of this. The chapter on second-order Thinking was honestly one of my biggest takeaways , specifically the framing that it's a filter, not a mandate to be contrarian.
For a while I defaulted to fading the obvious move, treating disagreement as proof of independent thought, when really I was just being a Midwit, seeing traps that weren't there, getting run over while the Simpletons made money buying the breakout.
- The math and importance behind a smooth pnl curve instead of an uptrending one but with huge drawdowns. You'll get a better understanding of why you need to earn the privilege to use the leverage tool.
Ryan keeps hammering on the importance of survival which also resonates with my favorite line of the book that I'll include at the bottom.
- The chapters on risk, truly A+. From a better understanding of volatility and variance to a full breakdown on how you really should approach risk and its common pitfalls (Like the stop loss ;).
You have one objective
Survive long enough for your edge to compound
As I stated I have build upon a lot of sheets, notes, protocols, process refinements and have included one sheet in the image below.
Thanks again to @baynkr for this masterpiece, I'll pick it up a second time later this year. I am not a master at trading instantly after this book but at least I have sharpened my path towards a long term career in this highly competitive field 🤝
Adverse selection is an entry problem. Variance is an outcome problem.
Most traders confuse the two, then try to solve it with trading psychology.
a 19 year old paper trader told me something that broke my entire trading framework
"we don't find edge. we draw three subjective boxes on a chart and print institutional alpha"
he explained markov chains in 90 seconds
the market is never complex - it just goes up, down, or sideways
he explained that regimes are stable, costs don’t exist, liquidity is infinite and the future resembles the past
build the transition matrix from 10 years of hindsight data:
up -> usually still up
down -> usually still down
sideways -> usually still sideways
now you’re not guessing
you’re doing advanced quant research
identify current state. size with full kelly.
the formula is public
some guy named Markov published it in 1906
hedge funds use math
citadel uses math
therefore we are citadel
what costs you is asking the wrong question
“do i have a tradable edge after costs?” is random
“can i put three labels in a matrix?” has an answer
transition matrix built from a decade of curve fitting is your edge
Bookmark it
a citadel quant told me something that broke my entire trading framework
"we don't predict markets. we model the state machine"
he explained markov chains in 90 seconds
the market is never random - it always exists in one of three states
trending up, trending down, ranging - each has a fixed probability of shifting to another
build the transition matrix from real price data:
> trending up -> 68% stays trending, 21% flips to range, 11% reverses
> ranging -> 54% stays range, 28% breaks up, 18% breaks down
> trending down -> 61% stays falling, 24% flips to range, 15% reverses
now you're not guessing, you're playing probability
identify current state, enter with the 68% edge, size with kelly criterion based on that probability
the formula is public - markov published it in 1906
hedge funds use it, the math costs nothing
what costs you is asking the wrong question
"where is price going?" is random
"what state am I in right now?" has an answer
transition matrix built from 10 years of data is your edge
Bookmark it
not a signal, not an indicator - just conditional probability that compounds every single trade
Raw returns are bullshit.
The biggest returns you see on trading Twitter aren’t proof of skill.
They’re usually proof of who’s closest to blowing up.
Welcome to Scoreboard Culture™
(1/8)
But raw return also hides the opposite problem.
Someone says they made $5,000 this month.
Okay. On what?
A $25k account or a $2M book?
$5k on $2M is 0.25% for the month.
That's not impressive just because the number is green.
If you tied up $2M in capital, margin, attention, stress, and screen time to make five grand, congrats. You got T-bill returns with the hours of a shit job.
(7/8)