Onur
1.7K posts
Onur
@OnurES_
building agents | prev data & ml @opensea @nansen_ai @mcgillu
Toronto Katılım Nisan 2018
3.9K Takip Edilen1.9K Takipçiler
if you haven’t tried /goal, you should.
@OpenAI cooked
and gpt5.5 is just amazing model to pair it with
Onur@OnurES_
/goal is amazing
English
Onur retweetledi
this aged so well so far
will age even better
Andrej Karpathy@karpathy
The hottest new programming language is English
English
Onur retweetledi
JUST IN: Accenture to roll out Microsoft Copilot to 743,000 employees.
English
i love codex with 5.5, it just one shots everything. great job @OpenAI @thsottiaux
English
My best ad creative is getting clicks. But no installs.
USD$0.41 per click
Best I've achieved so far.
But no installs.
It tells me this is hooking into a need. They click. But then the app store listing itself doesn't deliver. So they don't install.
Adam Lyttle@adamlyttleapps
I’ve been A/B testing meta enhancements (their ai making “improvements” to my creative to increase conversions) And… I don’t think it works 😂
English
Same experience.
I have been working on a migration project and made some progress with 5.4 but it got stuck, was running into the same issues.
5.5 medium thinking just crushed it
banteg@banteg
i have a reverse engineering project i started with 5.3-codex. after a few days it hit a wall. then i had meaningful progress when 5.4 came out. but the original code seems quite convoluted and cursed, so it got bogged again. now 5.5 is breezing through it again. it's pretty interesting that there is a tangible wall in understanding for each model.
English
who could have guessed these would generate less money than a kebap shop in berlin?
DBCrypto@DBCrypt0
Raised 9 figures Generates less revenue than a kid's lemonade stand
English
i think i was too bearish with this calculation. average employee will be using a lot more ai than 500/month.
if you're only using 500/month worth of ai it probably means that you could be replaced
a good benchmark companies can use (i know it's not perfect and gameable) is median ai spent per employee and when this drops too much it kinda indicates you can do another set of layoffs
Onur@OnurES_
if we assume ai will be a must at work going forward for not just software devs but for all white collar employees, assume only 1/3rd of US enterprises pay for ai and assume ai companies charge 500$/month per person you get 200b/year ai revenue already just from us, just enterprise everything else like personal ai assistants, ai devices ai generated entertainment (movies, songs, games) oai ads ai driven r&d idk sounds like eventually they will make a lot of money as long as ai progress continues
English
Onur retweetledi
Introducing Claude Design by Anthropic Labs: make prototypes, slides, and one-pagers by talking to Claude.
Powered by Claude Opus 4.7, our most capable vision model. Available in research preview on the Pro, Max, Team, and Enterprise plans, rolling out throughout the day.
English
i think we’re past this because a few things are v obvious now:
1 ai labs will have more than 100b arr are by the end of 2026 and more than 200b arr by the end of 2027
2 progress is not slowing
3 the capabilities are increasing and these things that started as chatbots have become systems that do actual work
4 frontiers models are now capable enough to create very serious cybersecurity risk. so having the most capable ai model is a national security priority and governments are waking up
for the investors this means you won’t get ai bubble blow up dip or panic sells because there is no bubble
ai labs will have enough revenue to pay for all the compute and more
Onur@OnurES_
at some point there will be a big selloff because people will think ai is a bubble, doesn’t bring revenue, progress has slowed, or it doesn’t need as much compute when that happens, buy ai stocks with everything you have
English
any reliable way of a codex session to launch multiple codex reviews in parallel within the session without stopping?
i want to see codex building and running reviews after small scoped task without me launching a codex review via slash command
i am sure there are multiple ways of doing this (i am doing this using cli) but perhaps there is a better way?
English
i often think about this
roon@tszzl
not enough people are emotionally prepared for if it’s not a bubble
English
Onur retweetledi