Ryan Brewer

819 posts

Ryan Brewer banner
Ryan Brewer

Ryan Brewer

@ryanbrewer

building finance things @openai

NY | SF Katılım Aralık 2018
785 Takip Edilen3K Takipçiler
Sabitlenmiş Tweet
Ryan Brewer
Ryan Brewer@ryanbrewer·
Excited to share that I’ve joined OpenAI to help build the future of financial intelligence. My focus is connecting models directly to the data sources, tools, and workflows analysts use every day. My team’s first product is the Excel plugin. After software engineering, finance will see the benefits of model improvements more acutely than almost any other field. This is just the start for us! More to come soon!
OpenAI@OpenAI

GPT-5.4 Thinking and GPT-5.4 Pro are rolling out now in ChatGPT. GPT-5.4 is also now available in the API and Codex. GPT-5.4 brings our advances in reasoning, coding, and agentic workflows into one frontier model.

English
61
46
931
242.3K
Ryan Brewer retweetledi
Tibo
Tibo@thsottiaux·
Codex will take us places
English
111
23
890
50.4K
bone
bone@boneGPT·
imagine paying 6.5 BILLION DOLLARS to hire Jony Ive only to pivot away from hardware without launching a single product gotta know when to fold em but damn 6.5 BILLION
bone tweet media
English
237
97
4.3K
898.3K
David Cramer
David Cramer@zeeg·
@ryanbrewer to be clear i think folks are getting value out of it, but the majority of this shit just looks like grift
English
1
0
1
34
Ryan Brewer
Ryan Brewer@ryanbrewer·
Do you think the models have gotten better at generating useful code over the last 6 months? Do you think you can look at that slope and see a trend towards beautifully written, production level code? Curious how you’re thinking about LLM progress plateauing, and what you’ve seen that shows this
English
1
0
0
72
David Cramer
David Cramer@zeeg·
@ryanbrewer i dont see any data suggesting that to be true I just see a lot of grift
English
2
0
5
200
Ryan Brewer
Ryan Brewer@ryanbrewer·
@zeeg Do you actually think this will be forever? IMO the plastic will default to gold very soon
English
1
0
0
209
David Cramer
David Cramer@zeeg·
"but [insert random absolute bullshit based no evidence whatsoever]" if you could do this, even 10k a day, you'd be able to replace almost any technology in the world except you cant, beacuse its absolute slop, you dont know what the software is doing, you didnt design any of it, it doesnt function, its unmaintainable this will forever now be known as plastic software
English
12
0
178
7.4K
David Cramer
David Cramer@zeeg·
where's all those billion dollar businesses built from gas town and other slop farms? oh
English
26
9
359
43.7K
Ryan Brewer
Ryan Brewer@ryanbrewer·
@_dmca When a measure becomes a target, it ceases to be a good measure
English
0
0
4
171
Ryan Brewer
Ryan Brewer@ryanbrewer·
@tszzl @_dmca When a measure becomes a target, it ceases to be a good measure
English
0
0
3
172
roon
roon@tszzl·
@_dmca honestly I’m becoming skeptical these are useful tokens
English
31
2
475
23.7K
Ryan Brewer
Ryan Brewer@ryanbrewer·
@gabemays @zeeg @sentry @zeeg has written some of the best software that the industry has ever seen. Sentry is a staple in almost any organization who takes themselves seriously. I’ve used it in every project and at every company I’ve ever worked at. I just disagree with the productivity part.
Ryan Brewer@ryanbrewer

Just have to get really good at managing a lot of context. Codebases are for agents now, it's your job to make sure the code they produce can scale! You should think about the code that's output as an increasing percentage of your prompt input tokens. If you allow models to write bad code to begin with, it become unmaintainable really quickly

English
1
0
0
34
Matt Bagni
Matt Bagni@mattbag00·
@ryanbrewer They slow us down a lot since we work in small teams and prs will need to be done more, longer, and across teams. The code productivity won't happen anytime soon
English
1
0
0
103
Ryan Brewer retweetledi
gian
gian@giansegato·
> hey claude i'm starting to work with this new team, what are the most important papers published in the last couple of years i should get familiar with > nice these look good. download them, put them in ~/downloads, and then print them. use the printer on my floor > [hearing the printer humming while warming up] this form of human-computer interaction is literally my childhood dream coming true. the reason why i started coding two decades ago. this is magic, i'll never get used to it
English
17
18
753
43.7K
bone
bone@boneGPT·
@spencerpapay @ryanbrewer They say a lot of things. They said stargate would happen with Trump at the White House. No credibility left. It will be folded too.
English
1
0
3
1.1K
Ryan Brewer
Ryan Brewer@ryanbrewer·
Again I’d point to the tweet I linked. There were always startups that wrote bad code, and they usually failed. It’s the same now! It’s on you to be able to use these tools to accelerate, and usually requires skilled engineering to receive the benefits (not writing unmaintainable slop). For my team and most teams @OpenAI I’d say this has been a massive acceleration in productivity
English
1
0
0
152
Alex
Alex@FullyAlexD·
Net-productive as a founder, yes. I built my fintech startup mostly manually early 2025. Was able to replicate the design patterns to build quicker. Startup was greenfield. Under 10k MAU. However, founders ive met on the otherside of the bottleneck are now experiencing the pains from going hands off without guardrails too soon. Re-writes, significant perf issues, churn from loss of trust. Productivity isnt measured by just output alone. You need to measure the tail where the economic benefit is actually measured.
English
2
0
0
164
Ryan Brewer
Ryan Brewer@ryanbrewer·
@FullyAlexD Are you more or less productive now than you were one year ago?
English
1
0
0
153
Alex
Alex@FullyAlexD·
@ryanbrewer "Net positive" is doing the heavy lifting in his statement. He even addresses your quoted response in his follow up comments.
English
1
0
4
152
Dmitry Shevelenko
Dmitry Shevelenko@dmitry140·
no company founded >2026 will organically grow to >10k employees
English
27
6
97
18.5K
Ryan Brewer
Ryan Brewer@ryanbrewer·
Slop buildup is real, but honestly something you have to dedicate a large portion of your engineering hours to. I focus mainly on making sure agents write code that I would be happy with. This compounds over time! If you allow codex to just roam free, eventually you will hit a wall. The art is in managing the code that's written, just like you would if you were a normal EM!
English
0
0
1
23
Benjamin Shafii | OpenWork (YC P26)
Benjamin Shafii | OpenWork (YC P26)@benjaminshafii·
the keyword is "net". there's 100% a ginormous initial boost on a new product. but the slop builds up and it becomes more and more unmaintainable. e.g. i built the current product in 24h, but i couldn't keep building at that speed afterwards. there were so many issues that crept up because of my incomplete mental model of the underlying code base. it requires new engineering mindset and techniques to reliably harness the boost of llms and i believe most people just haven't figured out how to do it yet.
English
2
0
4
224

Keşfet

@boneGPT @sama @zeeg @_dmca @tszzl @gabemays @sentry @elonmusk