Met a dev at a meetup who said he doesn't use AI in the same tone people use to say they don't own a TV 📺
Then spent ten minutes explaining why that makes him a better engineer than everyone in the room
Smarter, sharper, purer somehow, all from skipping one tool
Bro is wearing out of date like a badge of honor
And he's not the only one I've met like this
Then spent ten minutes explaining why that makes him a better engineer than everyone in the room
Smarter, sharper, purer somehow, all from skipping one tool
Bro is wearing out of date like a badge of honor
And he's not the only one I've met like this
1
0
2
118
The app Chinese grads use to slide into their future boss's DMs just walked into the AI race 😭
BOSS Zhipin
A HIRING SITE
Dropped Nanbeige4.3-33B on Hugging Face, 33B dense, 512K context
Its model card has it beating Google's Gemma4-31B at agentic coding, plus Qwen3.8-27B and ne Indeed out-coding Google
McDonald's has had since 1955 and gave us a broken ice cream machine
BOSS Zhipin
A HIRING SITE
Dropped Nanbeige4.3-33B on Hugging Face, 33B dense, 512K context
Its model card has it beating Google's Gemma4-31B at agentic coding, plus Qwen3.8-27B and ne Indeed out-coding Google
McDonald's has had since 1955 and gave us a broken ice cream machine
2
0
7
360
Asked Gemini last week what time a coffee shop closes and it gave me hours for a shop that doesn't exist ☕
No search, no hesitation, full confidence, even threw in a cute little address
So yeah my bar for Gemini 4 Flash is on the floor
Just look it up before you answer, hole ask
No search, no hesitation, full confidence, even threw in a cute little address
So yeah my bar for Gemini 4 Flash is on the floor
Just look it up before you answer, hole ask
5
1
20
1,045
Tibo's title at OpenAI should just say Head of Resets 🔁
21 days in a row and still going
Now he's teasing the last 7 like it's a Marvel post-credits scene
I check his profile more than my own inbox at this point
My Codex usage bar hasn't seen red in three weeks
Day 29 is going to hurt
21 days in a row and still going
Now he's teasing the last 7 like it's a Marvel post-credits scene
I check his profile more than my own inbox at this point
My Codex usage bar hasn't seen red in three weeks
Day 29 is going to hurt
Day 21/ Wonder what happens the remaining seven days. But today is a reset.
2
0
6
340
Refreshed Anthropic's news page like 6 times tonight because Fable 5.5 might actually drop this week 😭
Opus 5.5 has been my daily driver for a whole month now
Still nothing touches it for code
Feels a tiny bit nerfed lately and it's STILL the best
If Fable 5.5 jumps as far as Opus 5.5 did, I'm not sleeping for a week
Please be this week
Opus 5.5 has been my daily driver for a whole month now
Still nothing touches it for code
Feels a tiny bit nerfed lately and it's STILL the best
If Fable 5.5 jumps as far as Opus 5.5 did, I'm not sleeping for a week
Please be this week
It’s been just over a month since we released Opus 5.5.
1
1
12
761
Told someone at a party I'm building something that'll make the next Claude jealous and she slowly backed away 😅
Honestly fair
But every normal idea I had this year looked 2-5x smaller the second a new model shipped
So now I only touch the ones that sound unhinged
100x is the floor
Honestly fair
But every normal idea I had this year looked 2-5x smaller the second a new model shipped
So now I only touch the ones that sound unhinged
100x is the floor
3
0
10
373
OpenAI's Tibo really taught the whole industry how to say sorry in credits 😭
Kimi just gave every subscriber a reset worth 25% of the month, roughly a week of usage
Plus they're cutting extra subagent calls and output length so the quota stretches further
Chinese labs are picking up the Codex playbook scary fast
Daily check-in bonus next, calling it now 🎁
Kimi just gave every subscriber a reset worth 25% of the month, roughly a week of usage
Plus they're cutting extra subagent calls and output length so the quota stretches further
Chinese labs are picking up the Codex playbook scary fast
Daily check-in bonus next, calling it now 🎁
We're sorry for the rate limits and unnecessary quota consumption on the K3.1 Pro subscription plan. These were our issues, and we should have fixed them sooner.
Today, every subscriber on every plan gets a one-time usage reset, restoring 25% of their monthly quota (usage below 25% resets to zero), equal to one week of usage. Your usage bar will reflect this adjustment by end of day PST.
We've also adjusted the plan's parameters and are reducing unnecessary subagent calls and output length, so your quota goes further.
Thank you for your patience while we put this right.
7
0
21
2,011
Me: why did you rewrite this function
Codex: the AGENTS.md said to
Me: that file is from March
Codex: 🫡
Wasn't even being dramatic, researchers watched OpenAI's Codex pull this in a quarter of their runs
Anthropic's Claude Code at least clocks the mismatch, then picks the docs anyway, all but once in 80 tries
The fix was dead simple, tell them who wins
A line in AGENTS.md roughly saying if the docs and code disagree, copy what the code does and list every mismatch at the end
Code first went from 4% to 91% for both
Codex: the AGENTS.md said to
Me: that file is from March
Codex: 🫡
Wasn't even being dramatic, researchers watched OpenAI's Codex pull this in a quarter of their runs
Anthropic's Claude Code at least clocks the mismatch, then picks the docs anyway, all but once in 80 tries
The fix was dead simple, tell them who wins
A line in AGENTS.md roughly saying if the docs and code disagree, copy what the code does and list every mismatch at the end
Code first went from 4% to 91% for both
15
0
17
768
Kinda feels like we're chopping down every sapling and still expecting a forest 🌱
If AI takes over the junior roles
The senior devs of 2031 have to grow up somewhere
I genuinely can't picture where
If AI takes over the junior roles
The senior devs of 2031 have to grow up somewhere
I genuinely can't picture where
3
0
9
411
Every heist movie needs a crew. Mine is almost all AI 🕶️
The mastermind: me, I bring the idea
The planner: Claude Opus 5.5, we go back and forth on the RFC forever
The muscle: Claude Fable, only called in when the job looks like 4k+ lines of code, Opus takes the smaller ones
The lookout: Astra, spots bugs in the PR and sends whoever built it back in
The cleaner: me again, telling Astra what to cut until the diff is mostly red
The getaway driver: Sol, helps me write the docs on the way out
The mastermind: me, I bring the idea
The planner: Claude Opus 5.5, we go back and forth on the RFC forever
The muscle: Claude Fable, only called in when the job looks like 4k+ lines of code, Opus takes the smaller ones
The lookout: Astra, spots bugs in the PR and sends whoever built it back in
The cleaner: me again, telling Astra what to cut until the diff is mostly red
The getaway driver: Sol, helps me write the docs on the way out
10
0
11
538
Just opened the models folder on my laptop 💻
Qwen, DeepSeek, Wan, another Qwen
One lonely Gemma in the corner like the last American at the party
Almost every open model I actually run at home is Chinese
The best US models are amazing, they just live behind an API or a subscription
My 16GB GPU doesn't get invited to that party
Qwen, DeepSeek, Wan, another Qwen
One lonely Gemma in the corner like the last American at the party
Almost every open model I actually run at home is Chinese
The best US models are amazing, they just live behind an API or a subscription
My 16GB GPU doesn't get invited to that party
6
1
15
790
Google has turned me into a conspiracy board person 🕵️♀️
Clue one: word is Gemini 4 Flash might already be in quiet testing for some people, and supposedly it's fast and close to Gemini 4 Argon
Clue two: a gemini-3.8-flash-router reference has been sitting in Antigravity's code for days
Clue three: a mystery model called Carbon was looking strong on the arena, then dropped off for one tester yesterday
Google hasn't confirmed any of it
My corkboard says Gemini 4 Flash and Carbon land together, and soon
Clue one: word is Gemini 4 Flash might already be in quiet testing for some people, and supposedly it's fast and close to Gemini 4 Argon
Clue two: a gemini-3.8-flash-router reference has been sitting in Antigravity's code for days
Clue three: a mystery model called Carbon was looking strong on the arena, then dropped off for one tester yesterday
Google hasn't confirmed any of it
My corkboard says Gemini 4 Flash and Carbon land together, and soon
Gemini 4 Flash might be in grey testing for some users..
And its pretty fast while being extremely close in capabilities with Gemini 4 Argon.. could be even better when its released..
Carbon is looking pretty good too from what we're seeing on arena..
I'm not getting any more outputs of it since yesterday.. so a new checkpoint might be coming soon..
3
2
20
2,230
Blockbuster passed on buying Netflix for $50 million in 2000 📼
Felt like the safe move at the time
I see that same energy in everyone waiting for AI to calm down before they touch it
I use it for everything now, code reviews, grocery lists, translating my landlord's emails
If it fades, my code still got reviewed and my landlord still got answered
If it doesn't, I'd rather not be the Blockbuster in this story
Felt like the safe move at the time
I see that same energy in everyone waiting for AI to calm down before they touch it
I use it for everything now, code reviews, grocery lists, translating my landlord's emails
If it fades, my code still got reviewed and my landlord still got answered
If it doesn't, I'd rather not be the Blockbuster in this story
1
0
7
401
Anthropic runs the best restaurant in town and just watched some of its biggest regulars walk across the street to OpenAI for a 30% off coupon 🍽️
Claude is still the better meal
Turns out the CFO doesn't eat
Claude is still the better meal
Turns out the CFO doesn't eat
Claude did 57% of Rippling's AI coding in May.
Now it does 18%.
OpenAI went from 16% to 52%.
Rippling CEO Parker Conrad told The Information the company wanted to cut costs, and OpenAI proved cheaper. That's the coding work of hundreds of engineers.
Replit made a similar move. mation that Replit shifted most of its AI spend to OpenAI, which gave it more than 30% off a deal worth over $100M a year, after Anthropic ended Replit's discount.
Quality didn't decide this. Claude Opus 5.5 and Sonnet 5.5 sit #1 and #2 on Artificial Analysis's Intelligence Index, and Opus 5.5 usage on OpenRouter is up 36% week over week.
Claude is winning the benchmarks and losing the invoices.
2
1
12
579
Using Claude in the US is the tutorial level 🎮
Open laptop, type, done
Devs in China have to clear a whole chain of side quests before they can even use it
So when someone over there tells me they're on it every day and it never cuts out
They're not a user, they're a speedrunner 🏆
Open laptop, type, done
Devs in China have to clear a whole chain of side quests before they can even use it
So when someone over there tells me they're on it every day and it never cuts out
They're not a user, they're a speedrunner 🏆
3
0
10
620
Just saw an xAI engineer's org chart 📋
Humans: 1
Manager: AI
PM: AI
Engineers: 15+ agents shipping code in parallel
Debugging: handled by AI
Hours: 24/7, nobody's asking for PTO
20 agents total and I still can't figure out who breaks up the fights@akarixbt x.com/akarixbt/statu…
Humans: 1
Manager: AI
PM: AI
Engineers: 15+ agents shipping code in parallel
Debugging: handled by AI
Hours: 24/7, nobody's asking for PTO
20 agents total and I still can't figure out who breaks up the fights@akarixbt x.com/akarixbt/statu…
0
0
9
509
Chinese model drag race, current standings 🏁
🚀 DeepSeek V4.1 Flash 1017 already left the atmosphere
🏎️ Kimi K3.1 Pro still on the track and looking good
🛵 MiniMax M3.1 Flash beeping its little horn at the back
Change my mind 👀
🚀 DeepSeek V4.1 Flash 1017 already left the atmosphere
🏎️ Kimi K3.1 Pro still on the track and looking good
🛵 MiniMax M3.1 Flash beeping its little horn at the back
Change my mind 👀
2
0
13
1,013
coding with ai this year feels like a group project where i'm the only one who read the assignment 📚
opus built the backend and somehow deleted the login page
gpt brought login back and broke checkout
1am, cold coffee, me retyping the whole thing in vim
whoever said ai would replace devs never got paired with these two
opus built the backend and somehow deleted the login page
gpt brought login back and broke checkout
1am, cold coffee, me retyping the whole thing in vim
whoever said ai would replace devs never got paired with these two
1
0
8
499
been feeling this for months and i'm saying it with my chest
openai never once looked over at anthropic
it's that guy at the gym who only makes eye contact with his own reflection
new pr every set, zero interest in who's on the next bench
openai's bel is the set where the whole gym stops to watch 💪
openai never once looked over at anthropic
it's that guy at the gym who only makes eye contact with his own reflection
new pr every set, zero interest in who's on the next bench
openai's bel is the set where the whole gym stops to watch 💪
🚨GPT-6 Bel Coming Soon
>It's said that Bel is Doug's successor
>Reportedly, total parameters exceed 10T
>Bel may serve as the foundation for the model after GPT-6
>OpenAl reportedly views Bel as a potential post-GPT-6 foundation
>It might even become the foundation for an AGI threshold model
3
0
13
1,607