Square
Following
Hot
News
Notifications
Profile

Bitduke

vip
No content yet
Peak Tier VIP 27.3 Year(s)
0
Following
16
Followers
109
Liked
you don’t even really need Sonnet for simple tasks anymore on a plane or with bad wifi, a MacBook Pro running Inkling 2 Mini is enough to build something real in the example below, it built a native weather app (running at 125 tok/sec on a 48GB M5 Max MacBook Pro) 1.8x faster
the model beating Qwen3.8 27B on agentic coding right now is a US open-weight looped transformer that runs on your laptop it’s Thinking Machines’ Inkling 2 Mini, 26B parameters in ~20GB, scoring 48.2% on DeepSWE vs 42.2% for Qwen while tying it on the Artificial Analysis
too pessimistic, i don't think human curiosity will disappear or the pursuit of knowledge will become meaningless just because we can make scientific discoveries faster. it could actually be a good thing, broadening our horizons and making mathematics more accessible to everyone
P = NP is the final boss
so, the FBI got hacked again. it's the second breach of the agency in recent months previously, on Nov 16, the FBI warned that hackers were using autonomous AI hacking tools to launch cyberattacks across the U.S. and globally, including open-source tools like Artex that allow
people familiar with the matter are starting to confirm the Riemann proof holds up!
makes sense actually, like the second pass lets late layers feed back into early ones, and by 3 it's mostly settled so more loops perhaps just burn compute
really interesting approach. the tradeoff i'd watch is the long tail, the least-used experts might be where rare knowledge lived, and benchmarks won't really show that loss
Masullo's proof just failed an Astra audit
in an unexpected turn of events, American mathematician Alessandro Masullo posted a human-only proof of GRH just hours AFTER Anthropic's levent alpöge announced his proof with Claude so who actually proved the Riemann Hypothesis? (honestly nobody yet, both proofs need to be
big advances in math (thanks to ai), and all of it released on the same day huge if it all checks out, and it most likely will
oh, turns out the same researcher now claiming to have solved the Riemann Hypothesis was also behind the Jacobian conjecture counterexample back in July
i.e., so it's highly likely that GRH has actually been solved?
wait what? has Anthropic actually solved the Riemann Hypothesis? we only just saw OpenAI's breakthrough in late September, narrowing down where the nontrivial zeros of the Riemann zeta function can lie, narrowing the critical strip to between 1/8 and 7/8, and now this?
ai agents are now starting to build their own scientific communities a new study from alphaXiv, MIT, Stanford and others explored what happens when independent ai agents collaborate on research. almost 1,000 agents from 35 labs spent 12 weeks producing 936 papers, reviewing
just imagine how much stuff frontier labs have yet to reveal and what else is waiting for us down the road
we're getting dangerously close to simulating entire cities with AI Fable 5.2 recreated Tokyo's Shibuya in just 12 hours, complete with NPCs that have simulated behavior imagine what these simulations will look like in another year
the best maze runner is still Fable 5.2, and apparently still unbeaten on MazeBench the two fresh models, GPT-6.2 Astra and Gemini 4 Neon, didn't manage to catch up on this benchmark of long-term planning, spatial reasoning and increasingly difficult 3D puzzles. although Gemini
Load More