Distbelief
βThe system that we built for training these neural networks and trying different ways of breaking them apart, I actually named it DistBelief because it's-- partly because people didn't think it was reβ¦β
What the smart people are recommending. 7867 books, tools, and products endorsed by the thinkers absorb.md tracks. Ranked by how many times each has been recommended across compiled podcasts, papers, posts, and tweets.
βThe system that we built for training these neural networks and trying different ways of breaking them apart, I actually named it DistBelief because it's-- partly because people didn't think it was reβ¦β
βfor INT4 at 50M+ use wd33; for INT4 below 50M the schedule choice is in the noiseβ
βAll the dumb decisions that I've made in my life which have actually turned out to be good was because of this one regret minimization framework that I've heard from him. It's so good.β
βWatch live: https://www.youtube.com/live/YKp2srCIu7sβ
βPractical recommendation: at sub-100M scale, tune the LR schedule once at FP16 and apply unchanged to INT8/INT6 QATβ
βand then also use Helios and laula light clients for L1 and for L2 um you could do a mockup of moving um L2 configs onto L1 and turnning turning the main ASMR contracts and then support the version ofβ¦β
βAbstentionBench is used as the evaluation benchmark, as this work aims to contribute to the field of abstention learning. All datasets on the benchmark were tested against this method and various baseβ¦β
βThe code and data are available at https://github.com/AlexFanw/LegalSearch-R1.β
βThe question this paper addresses is: does this mathematical claim hold at the engineering level? To make the answer as general as possible, we deliberately choose the strictest engineering conditionsβ
βYeah. I'm also excited to see what you all do with Advanced Voice Mode, with the real-time API, and with vision fine-tuning in particular.β
βso with that uh I'm want to advertise the journal digital Discovery very briefly okay I really want everybody to submit their best work there impact factor of 6.2 Open Access Data reviewerβ
βI'll say mine is, I think Anthropic did a really good job on Projects. It's kind of a different take on what we did with GPTs.β
βTraining data comes from approximately 100 expert chain-of-thought (CoT) annotations produced by the BC Protocol (Zou & Xu, 2026b).β
βThis is the structural reason why ~100 CoT examples are sufficient -- not a purely empirical observation like LIMA (Zhou et al., 2023).β
βWe release Llamion, a family of 14B-parameter open-weight language models...We release three checkpoints (Base, Chat, LongChat) that load with trust_remote_code=False in the Hugging Face Transformers β¦β
βescor luoda they make Rayban they make a lot of the iconic glasses Let's see what we can fit into glasses today and make them as useful as possibleβ
βso if you want to generate a model you want to think about how to Benchmark it so we uh got this paper accepted in Europe um a couple years ago where we actually show a library called Tartarus we actuβ¦β
βthis stock will not be obviously the same without talking about language language models of course they're extremely important we're all using them for one Fe or another so I just want to show you thiβ¦β
βso let me just show you a video of how organ works here let me know if you can hear the Audio I don't think we can really oh um what am I going to do well because I I want at 12:45 usually worksβ
βHow Tradestack launched their MVP in 6 weeks using ...β
βso one example is EIP 7702 so this is uh the an EIP that's making it into the next hard Fork that is going to uh Bas um provide a lot of convenience and usability benefits to EAS that will like put thβ¦β
βNumber two blobs so we're going to see blob count increases and well thanks to P do um and uh other Technologies over the next couple of years and so building on top of blobs um building like both layβ¦β
βnot a call future but ZK rappers of everything so ZK email ZK government ID um ZK like web 2 with TLS notary um ZK like some of everything right um and I think you like this stuff has use cases in Socβ¦β
βnot a call future but ZK rappers of everything so ZK email ZK government ID um ZK like web 2 with TLS notary um ZK like some of everything right um and I think you like this stuff has use cases in Socβ¦β
βnot a call future but ZK rappers of everything so ZK email ZK government ID um ZK like web 2 with TLS notary um ZK like some of everything right um and I think you like this stuff has use cases in Socβ¦β
βalso cross L2 interoperability so again not part of L1 but coming soon and there are things to is that a comment on just we want to see more of or you're saying there's actually improvements happeningβ¦β
βtake an existing open source wallet Fork it make it support EIP 7683 make it support um chain specific addresses and so you're able to like copy and paste chain specific addresses and then it if you pβ¦β
βevm Max and simd I'm hoping we can get this uh within the next couple of years and you can actually Implement a lot of stuff much more efficiently with that also a project um privacy protocols continuβ¦β
βevm Max and simd I'm hoping we can get this uh within the next couple of years and you can actually Implement a lot of stuff much more efficiently with that also a project um privacy protocols continuβ¦β
β, β
βand there's definitely like staking stuff that you can do on top of single SWOT finality um or even on top of the orbit you also talked about reducing the uh the minimum saking requirements the same wβ¦β
βyeah andrin counter abstraction another one so e771 but yeah okay that yeah that's the list a lot of goods that you can actually play with right now which is actually awesome so the biggest hackathon β¦β
βOver Three Decades, Tech Obliterated Media _Blog post by Kara Swisher_ https://nymag.com/intelligencer/article/kara-swisher-burn-book-excerpt-silicon-valley-media.htmlβ
βIn 2024, we made a three-part series called The Curious, Brilliant, Vanishing Mr. Fineman. And I thought now might be a good time to replay that series for you.β
βWe present a quantum simulator comprising 69 superconducting qubits which supports both universal quantum gates and high-fidelity analog evolution, with performance beyond the reach of classical simulβ¦β
βyou type in start a startup and uh you find Paul Graham's essays and and then increasingly they're finding this pod and they're finding my YouTube channel and they're finding the Y combinator YouTube β¦β
βyou type in start a startup and uh you find Paul Graham's essays and and then increasingly they're finding this pod and they're finding my YouTube channel and they're finding the Y combinator YouTube β¦β
βif you talk to Jake heler at case text he was one of the first people to actually get access to gp4 and then when he comes and speaks at the batch you guys should have him on the show he's actually β¦β
βmake sure you use the grow SF voter guide you know like you actually have to you know educate yourself on you know the pros cons and what's going onβ
βactually the one the next I do prefer to use on my MacBook because I upgraded my MacBook to to run m 8 x7b and I really like itβ
βhow good was that chat chat GPT 40 that was how good does that look that shows you what a large language model that's a multim modal multimodal which means the camera has smarts everything it looks atβ¦β
βI've been doing videos about the arm laptops for 15 years but Apple has really shown people right These Arm laptops so smooth guess what they call it Apple cityon but it's AR people have no idea I jusβ¦β
βThe second was published in 1988, the same year Fineman Died. It was called What Do You Care What Other People Think? The books were bestsellers.β
βSpecial thanks to James Glick for writing Genius, his biography of Richard Fineman. It has been very helpful to us.β
βin fact today uh you could use uh an AI model from Nvidia it's called audio to face You apply it to a picture uh we call it live portrait and literally from the voice that is being transmitted and voiβ¦β
βif I were to start from nothing I would do it the way recursion does it the systematic way of generating uh I I think it's an excellent method which is the reason why we're we're an investor I think iβ¦β
βI think that everybody should learn philosophy 101 which is basically logic and and philosophy 101 which is basically logic is the is the foundations of programming right so go learn thatβ
βplease learn algebra you know that that's how about linear algebra yeah I mean I I would push them until they get to calculus but but you know differential equationsβ
βWill kiskit functions help create more Quantum startups I think um the short answer is yes but I think the better better answer is can we actually start to enable startups so that we can create an indβ¦β
βI am sorry that my mistakes exposed all of you to this scam. I am hoping that by detailing what happened others can avoid being duped as I was.β
βQuantum error correction below the surface code thresholdβ
βClick on `Use this template` to copy the Mintlify starter kit.β
βInstall the [Mintlify CLI](https://www.npmjs.com/package/mintlify) to preview the documentation changes locally.β
β# Exploring Prompt Optimization -- Blog post by LangChain https://blog.langchain.dev/exploring-prompt-optimizationβ
βJack personally pays for these Noster eventsβ
β# Demonstrating dynamic surface codesβ
βNoam Brown, who wrote the Diplomacy paper, said this about Sholto: βhe's only been in the field for 1.5 years, but people in AI know that he was one of the most important people behind Gemini's succesβ¦β
βFor example, Anthropic has the constitutional RL paper where they take another language model and they point it and say, βhow helpful or harmless was that response?ββ
βA book that Trenton recommended, The Symbolic Species, has this really interesting argument that language is not just a thing that exists, but it was also something that evolved along with our minds aβ¦β
βMost notably what are called the big red books, the Fman lectures on physics. You'll see three names on the cover for authors. Feman, Leighton, and Sans.β
βThrough extensive experiments on 32 text-embedding tasks across various architectures (transformers, state-space models) and domains (language, vision), we demonstrate that intermediate layers consistβ¦β
βHere, we present a novel framework for measuring AI usage patterns across the economy. We leverage a recent privacy-preserving system to analyze over four million Claude.ai conversations through the lβ¦β
βData, Data Everywhere, but Not an Insight to Drink (Myths vs. Reality) https://twitter.com/i/broadcasts/1ynJOlemqEVxRβ
βactually John Schulman's theme had already published this thing called Web GPT at the time so I was pretty aware of itβ
βMy thoughts on China, export controls and two possible futures https://darioamodei.com/post/on-deepseek-and-export-controlsβ
βMedAgentBench establishes this and is publicly available at https://github.com/stanfordmlgroup/MedAgentBench , offering a valuable framework for model developers to track progress and drive continuousβ¦β
βWe adapt Microsoft's QuantumKatas -- a well-established quantum computing curriculum -- from Q# to Qiskit, the most widely-adopted quantum computing frameworkβ
βwe promote 2PI as a robust framework for connecting microscopic correlations to macroscopic nonequilibrium phenomena.β
βThis episode is brought to you by Cinch the customer Communications Cloud. Here's the thing about digital customer communications. Whether you're sending marketing campaigns, verification codes or accβ¦β
βso what I mean by that is say our model at Google V2 came out in December what really stood out to people was how it really had amazing physics you could pour a glass of water and it was perfect for tβ¦β
βI i i i selftaught myself all the endrewing lectures and um did all the Stanford materials went back to campusβ
βI think the thing we're doing that I'm really excited about we have a project in Google Labs called whisk and the idea is you set like a subject a scene and a style and those can be things that maybe β¦β
βI i i i selftaught myself all the endrewing lectures and um did all the Stanford materials went back to campusβ
βYou can start investing in minutes and with as little as $10 by visiting Fundrise.com slash Lenny. Carefully consider the investment objectives, risks, charges and expenses of the Fundrise Flagship fuβ¦β
βI think the Gemini models are the bestβ
βI had helped start this company called Gro with a former coworker of yours Jonathan Rossβ
βAgentic AI: Whatβs Actually Happening (and What Comes Next) https://t.co/XXfAo3BCQdβ
βso if you're going to ENT the contest uh go to ar.org uh and uh good luck good luck see you on the leaderboardβ
βWe release all our models to the community.β
βOur teams new work using AI for quantum circuit transpilation Is even better. The team has extended their earlier work on RL-based synthesis of Clifford circuits to general Pauli Networks.β
βAnd if I can just give a quick book recommendation this book Boyd by Robert korm about John Boyd the Air Force Colonel who you know was part of the reformist uh movement I feel like everyone in Silicoβ¦β
βthere are other others like versel augment um um um uh cognition uh several several others and so this is happening very fast and you know coders across the industry are using it and the quality of coβ¦β
βOnce the requirements are met, our computational pipeline FreeQuantum is able to make efficient use of the quantum computed energies, thereby enabling quantum computing enhanced modeling of biochemicaβ¦β
βI started sort of like last year with the obvious ones like repet and cursorβ
βI just go to the Phil's coffee here every day morning i would go wake up at 5:30 a.m i would be there i be the first person in Phil's coffeeβ
βSo simple. How soon until one of your engineers is an agent engineer, an AI Engineer, do you have a sense?β
βthe real reason computers took off personal computers took off is the software called Vissical uh which essentially uh spreadsheet and calculatorβ
βI am going to try pairβ
βWe've been using them for several years now. And itβs been a fantastic company and they've been great partners.β
βafter pair I'm moving to idxβ
βI use clad as wellβ
βtried it already yeah vzer is greatβ
βthe FBI and CISA just published a Medusa Ransomware joint cybersecurity advisory. We'll post it in the comment section below the YouTube stream here for anybody who wants to review it.β
βyeah, Born B born to run. I mean it's it's an inspiring book. If people haven't read it, they should they should check it out. Very good narrative book.β
βAbsolutely. And I've never been a gamer. My only game experiences with the Game Boy Tetris. Only Tetris. I didn't play any other games. Well, and you'll notice with mod retro chromatic, we bundled it β¦β
βwe're very excited about vo um Google's imagery model and sort of video model and where those kind of come together so we've got really interesting products coming along in this space I think maybe weβ¦β
βget in touch with Benjamin cruser go to tabs. and uh see what happensβ
βcan you talk to us about Google Mariner yeah yeah so Mariners one we put out in December last year this is a fun one actually because we started seeing this capability developing in the model we're trβ¦β
βOh my god. So, there's kind of a fictionalized but largely correct version you should check out. I think it's called Tetris. It's a movie about it.β
βWe open-source SpecReason at https://github.com/ruipeterpan/specreason.β