Your RTX 3090 runs some Tensor Core math at half Nvidia's pro-card rate.
gpu-unlock, an open-source Linux kernel module, lifts that cap before Nvidia's driver loads. No VBIOS flash. One report: BF16 from 77.5 to 129.6 TFLOPS.
Catch: token generation barely moves. Prefill does.
gpu-unlock, an open-source Linux kernel module, lifts that cap before Nvidia's driver loads. No VBIOS flash. One report: BF16 from 77.5 to 129.6 TFLOPS.
Catch: token generation barely moves. Prefill does.
0
0
0
23
Two Claude Code builders left Anthropic in one day.
Anthony Morris, who created Claude Code on desktop: "Yesterday was my last day." Thariq posted his own exit hours earlier.
Neither gave a reason. Both land weeks before Anthropic's reported IPO.
x.com/amorriscode/st…
Anthony Morris, who created Claude Code on desktop: "Yesterday was my last day." Thariq posted his own exit hours earlier.
Neither gave a reason. Both land weeks before Anthropic's reported IPO.
x.com/amorriscode/st…
Yesterday was my last day at Anthropic.
I created Claude Code on desktop because I believe everyone can be a builder. Every feature request and bug report we got from this community was a gift that impacted millions of lives. I’m eternally grateful for those gifts, the product wouldn’t be what it is without you.
Anthropic is doing some of the most important work of our time. Contributing to that mission will forever be a career highlight for me. I never imagined I’d get to contribute to something so special in my lifetime.
Walking away from my best friends, a product I poured my heart into, and the mission was the hardest thing I’ve done in my career. But now I have some free time for GTA VI.
I'm already itching to get back to addressing all your feedback. I'll share what's next soon!
0
0
0
35
"No human can write better code than LLMs."
Xiangpeng Hao on bbtree, a Rust B-tree where LLMs wrote all the code. His benchmarks: 5x faster than std's BTreeMap on one thread, 12x faster than crossbeam-skiplist at 96 threads.
Speed is easy. Proving it correct is the hard part.
Xiangpeng Hao on bbtree, a Rust B-tree where LLMs wrote all the code. His benchmarks: 5x faster than std's BTreeMap on one thread, 12x faster than crossbeam-skiplist at 96 threads.
Speed is easy. Proving it correct is the hard part.
0
0
0
11
My read: by mid-2027, every major AI assistant makes phone calls for you.
Grok Bot can now call a person or business, make your request, take notes and send you a summary.
The rest follow fast.
Which call would you hand off first?
x.com/elonmusk/statu…
Grok Bot can now call a person or business, make your request, take notes and send you a summary.
The rest follow fast.
Which call would you hand off first?
x.com/elonmusk/statu…
Grok @bot can now call someone or a business for you, make a request on your behalf, take notes on the response and send you a note summarizing the conversation!
0
0
0
8
Why does AI writing feel harder to read? A new paper measured it.
All nine frontier models tested use more rare phrases and odd constructions than humans. The authors' hypothesis: models are trained for math, code and reasoning, which rewards density, not clarity.
Noticed it?
All nine frontier models tested use more rare phrases and odd constructions than humans. The authors' hypothesis: models are trained for math, code and reasoning, which rewards density, not clarity.
Noticed it?
0
0
0
3
Two engineers applied for the same job.
One finished the take-home in 40 minutes with AI. Flawless.
The other took two days, wrote it by hand, and left a bug in it.
The team hired the second one. The reason started a fight in the hiring channel.
Candidate A's code was clean. Tests, types, docs. Better than half the team's code.
Candidate B's was plain, a little slow, and missing a check.
So in the live round, the interviewer asked one thing: "Why does this function retry three times?"
"That's what the model suggested. It's a common pattern."
"What happens if the service it calls is down for ten minutes?"
Long pause. Then: "I'd have to ask the model."
Candidate B's bug came up in the live round. B stared at it for 30 seconds, said "I'm not handling the empty list," fixed it, then flagged two edge cases nobody had asked about.
Half the panel said: A is the future. Don't hire for a world that's gone.
The other half said: A can generate code. B can debug it. When the agent's code breaks at 3 AM, you need someone who knows why.
They hired B. Then they changed the process. AI allowed in the take-home. The live round is all about explaining and breaking your own code.
A made-up scenario. But plenty of hiring panels are having this exact argument right now.
Who would you hire: A or B?
One finished the take-home in 40 minutes with AI. Flawless.
The other took two days, wrote it by hand, and left a bug in it.
The team hired the second one. The reason started a fight in the hiring channel.
Candidate A's code was clean. Tests, types, docs. Better than half the team's code.
Candidate B's was plain, a little slow, and missing a check.
So in the live round, the interviewer asked one thing: "Why does this function retry three times?"
"That's what the model suggested. It's a common pattern."
"What happens if the service it calls is down for ten minutes?"
Long pause. Then: "I'd have to ask the model."
Candidate B's bug came up in the live round. B stared at it for 30 seconds, said "I'm not handling the empty list," fixed it, then flagged two edge cases nobody had asked about.
Half the panel said: A is the future. Don't hire for a world that's gone.
The other half said: A can generate code. B can debug it. When the agent's code breaks at 3 AM, you need someone who knows why.
They hired B. Then they changed the process. AI allowed in the take-home. The live round is all about explaining and breaking your own code.
A made-up scenario. But plenty of hiring panels are having this exact argument right now.
Who would you hire: A or B?
0
0
0
22
