Aṇukalpānta | testing, exploring and working | discord guy | community team at @dawninternet | love cnn, ml and llm | training somewhere | web3 holder

science and technology
Joined September 2015
we know they happen cuz we can measure their effects but we cannot explain why the universe is built this way instead of another way)
1
12
1,427
company to company learn from it, u cant just fight if u dont have model that do what hf did
they've been non stop trying to attack our systems for the past 48 hours a side effect is everything is getting hardened but damn we basically need agents going 24/7 because that's what they have
47
lmao
First in my bloodline to use an API key
89
kya?
ユーザーのみなさまへ Gboard チームからの新しいご提案です。 キーが指先まで流れてくるキーボード 『Gboard くるくるバージョン』を開発しました。 フルバージョンはリプライでご確認ください↓
1
64
0.1mb noice👀
1/ Today we're releasing Lattice Jolt: a post-quantum version of the Jolt zkVM, built on lattices instead of elliptic curves. It's _faster_ than curve-based Jolt and has the shortest proofs of any post-quantum zkVM — under 100 KB. Post: a16zcrypto.com/posts/article…
1
111
in three and a half years we went from a system that could not reliably perform elementary arithmetic to a coordinated network of 10 thousand agents capable of producing a formal soln to one of the millennium prize problems
1
75
🔥huge
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics. The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra. The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
1
76
principle i have learned from last 5 months been training models (effectivity), is that intelligence isn't only a property of the model. its also a property of the environment in which the model is allowed to learn
3
62
if your boy: - stays online 24/7 - runs hot under pressure - never drops context that’s not your boy. thats a hyperscale rack
5
71
Hendrix.btc ⭕🔶 retweeted
The most interesting products I've fallen in love with this year don't really have an interface. You don't need to open another app, go through another onboarding, or find another dashboard. You simply reach it the way you'd reach a friend, through a text thread or a phone call, and on the other side sits an agent that is connected to all the tools you use every day, such as your email, your calendar, your messages, your accounts. Then you can simply tell it what you need and it goes off and does it for you. Replies drafted, trips arranged, subscriptions cancelled, appointments moved. The apps it operates are all still there, humming underneath, but you just never see them anymore. They disappeared. Watching this happen has clarified something I've been thinking about for a while. We are witnessing a new step being inserted in front of every interface we've ever designed. In the last years, using software meant going to the software, opening the app, learning its layout, and navigating its particular logic. Now you increasingly interact with one interface that is connected to all the other interfaces, and the moment that layer works well enough, the interfaces behind it begin to fade from view. They don't fade because they got worse, but because nobody is looking at them anymore. That is what an invisible interface actually means to me, and it has nothing to do with minimalism or fewer buttons. It means that the interface of the product you built, the thing your team obsessed over and tested and polished for years, basically becomes a back-end, and its primary user is no longer a person but an agent acting on a person's behalf. I would encourage every designer to sit with that thought for a moment, because it is genuinely destabilizing. The craft I grew up in assumed a human on the other side of the glass, and every decision we made about hierarchy, contrast, motion, or copy existed to help that person see, understand, and act. When the entity clicking through your product is an agent, most of that craft becomes invisible to the one doing the clicking, and your beautiful checkout flow gets traversed in two hundred milliseconds by something that has no appreciation for your easing curves. This doesn't mean design is no longer relevant, but it does mean it is migrating to another level. The interface layer of an individual service is turning into something closer to a protocol, which needs to be legible, structured, operable, and honest about its state. Designing for machines is still design, just in a different form. The human-facing craft, the part of our work that earns love and loyalty, moves up a level to wherever the person actually is, and increasingly the person is sitting in one conversation that touches everything. Full essay here: open.substack.com/pub/design… LFG
16
5
4
119
9,556
an idea is only potential until u give it form. putting pen to paper turns what exists in your imagination into something real something u can see and build upon. thats why i often return to the simplest tool we have: pen and paper))
2
3
104
maybe there is a reason pen and paper still feels so natural. the human mind evolved to understand the world by interacting with things we can see and move??
28
Hendrix.btc ⭕🔶 retweeted
Privacy is a human right Crazy how eager society is to dismantle it at every level
105
90
18
790
81,810
OAI named their first chip jalapeño so the competition would feel the heat xd
Since announcing Jalapeño, our first custom inference chip, we’ve been testing it and the system around it. The results show a major advance: more intelligence from every watt and faster responses, delivering both higher throughput and lower latency in one architecture without sacrificing efficiency.
2
111
Hendrix.btc ⭕🔶 retweeted
What is the simplest, cleanest technique for self-supervised learning (SSL)? RotNet (2018), without a doubt Take an image, rotate by either 0, 90, 180, 270 degrees, and predict its rotation. The model must learn useful features to solve the task. Read on... [1/2]
26
29
7
431
82,916
hey hey🤗 was reading this new RESTRAIN paper on self driven RL with self penalization tbh pretty wild how far label free reasoning can go now. nice short read if u have a bit of time ) arxiv.org/html/2510.02172v1
1
2
61
accept death man anxiety follows. master it and acceptance becomes real then u live fully
Roughly 2/3 of humanity believes in some form of life after death. Humans refuse to accept death.
1
92
its being said but still, kinda scary how much critical infrastructure still needs to be properly secured :/
75