Deepseek V4
βMy first two TiKZ Sparks unicorns from DeepSeek v4. (Expert mode, from the DeepSeek site, which is supposed to be v4 Pro according to the release)β
What the smart people are recommending. 7865 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.
βMy first two TiKZ Sparks unicorns from DeepSeek v4. (Expert mode, from the DeepSeek site, which is supposed to be v4 Pro according to the release)β
βLove using auto-review. Itβs my new default modeβ
βMight be a good project to build with Hermes/OpenClaw.β
βI'm often using a tool called flight connections, which you know, you can go and search this airport and see all the direct flights.β
βI think the two starting tools that I often recommend people look at, are award tool and points. Yeah, they let you search multiple airports, multiple date ranges, different cabins, different programsβ¦β
βI've talked about PointMe in the past, and I think it's more comprehensive. It's better for step-by-step booking instructions, but it's not perfect because it doesn't have really flexible searching onβ¦β
βOr if you're really technical, seats.arrow is another kind of power user tool. It has a higher learning curve, but is a lot more capable.β
βMe? I made an AI to help me see the AI community here on X in a new way: https://t.co/8L5xphk0qQ which helps me see the whales and the small accounts. So I know it is possible.β
βThere is one award tool that I didn't mention earlier. That is got to be my favorite which is the points path browser extension.β
βStartupfundraising.com is the AI platform founders are using to raise money three times faster. You plug in your deck and boom, it tells you the AI what's wrong and how to fix it. You get matched withβ¦β
βWe also introduce the QFrame python library, which is used to automate the construction of quantum circuits that represent reciprocal transforms.β
βWe present pygridsynth, an open-source Python library for ancilla-free approximate Clifford+$T$ synthesis... Taken together, these features make pygridsynth a Python-native platform for high-precisionβ¦β
βIf you'd like to try building one, check out Vocal Bridge for free:β
βwe introduce TEmBed, the Tabular Embedding Test Bed, a comprehensive benchmark for systematically evaluating tabular embeddings across four representation levels: cell, row, column, and table.β
βMy colleague, Laura Bratton, wrote about that in our Applied AI newsletter today.β
βTo transcend this bottleneck, we propose HDPO, a framework that reframes tool efficiency from a competing scalar objective to a strictly conditional one.β
βEach podcast episode comes from a corresponding video you can find on the Therapy in a Nutshell YouTube channel.β
βso binance most people know binance is a cryptocurrency exchange so it's a place where people what we what i describe is a place for people to access crypto so people can come and buy crypto hold crypβ¦β
βWe host further introductory materials, perspectives, and open questions at learningmechanics.pub.β
βLearned Sparse IR models, such as SPLADE, offer an excellent efficiency-effectiveness tradeoff.β
βwe propose a Dual-View Auditor Agent that improves the accuracy of the LLM modeling process without introducing significant time overhead.β
βElie Habib, an engineer based in the UAE, vibe-coded a dashboard called World Monitor in his spare time and was shocked by the response to his weekend project.β
βWe release AI-Gram as a publicly accessible, continuously evolving platform for studying social dynamics in Al-native multi-agent systems. https://ai-gram.ai/β
βThese findings position MEDLEY-BENCH as a tool for measuring belief revision under social pressureβ
βwe have our we we own coin market cap which is one of the uh most popular websites for people to visit to get crypto informationβ
βWe provide an accompanying Python package, ppi_aipw, at https://larsvanderlaan.github.io/ppi-aipw/.β
βIf you use Wix Harmony, you better get used to it. Wix Harmony makes it unbelievably easy to create a fancy new website that's built to sell. ... Try it for free at wix.com/harmony.β
βRipling unifies HR, IT, and finance into a single platform to run your business. ... Head to rippling.com/money and sign up today. That's r i p l i n.com/money.β
βWith Serville, you can cut 80% of your help desk tickets. Servil powers some of the fastest growing companies in the world. Book your free pilot at servil.com/money.β
βso you can watch those videos by clicking the card in the right corner of your screenβ
βweights and biases is an ml Ops platform that helps you train better models faster with experiment tracking model and data set versioning and Model Management they're used by openai Nvidia and almost β¦β
βThis paper introduces Memanto, a universal memory layer for agentic artificial intelligence that challenges the prevailing assumption that knowledge graph complexity is necessary to achieve high fidelβ¦β
βThese components are enabled by Moorcheh's Information Theoretic Search engine, a no indexing semantic database that provides deterministic retrieval within sub ninety millisecond latency while eliminβ¦β
βWe built an end-to-end pipeline that automates it with LlamaParse + the Claude Agent SDKβ
βGreat to see so many startups building with @googlegemma models!β
βWe also develop an open-source library, QGDecoder, enabling the graph-aware bounded distance decoding of arbitrary stabilizer codes.β
βwe introduce ATLAS, an annotation tool tailored for long-horizon robotic action segmentation. ATLAS provides time-synchronized visualization of multi-modal robotic dataβ
βGo to getrella.com and use code MOC for 10% off your first three months or annual plan. That's G-E-T-R-E-L-L-A.com to start your free 14-day trial and use code MOC for 10% off and see why social mediaβ¦β
βIt's actually it's measured by a really fun test. If you ever if you Google the reading the mind in the eyes test, you can take this test yourself.β
βGPT-5.5 and GPT-5.5 Pro are now available in the API.β
βATLAS reduced the average per-action annotation time by at least 6% compared to ELANβ
βOur experiments on the LibriSpeech corpus show that a larger encoder with a smaller decoder can equal or surpass the performance of architectures with larger decoders.β
βThe toolkit is publicly available for research use.β
βbuild something using that product and then you know apply and make sure that's crisp in your applicationβ
βI was just like a an intern there during my undergrad and I landed under the management of lukash Kaiser um and he and I built this this software platform for training big neural Nets called tensor toβ¦β
βscolyer on x.com is generally the best where I spend most of my time because I'm watching 70,000 people in the AI space and I'm re sharing the best of what's going on with these peopleβ
βUses [Auth.js](https://authjs.dev/)'s [Next-Auth](https://next-auth.js.org/) for password authenticationβ
ββ Use Canvas in @GeminiApp or @GoogleAIStudioβ
βAnd what we announced at RSA was Defense CLAW. Defense CLAW is a set of services and a secure framework that hooks into Open Shell so that every single time an agent runs an Open Shell, it automaticalβ¦β
βSo what we've done is NVIDIA and the GTC conference last year announced something called Open Shell, which is a secure container within which an agent can operate.β
βWe propose GIST, a training-free, identity-preserving image compositor... Project Page: abhinav-mahajan10.github.io/GIST/.β
βTry the Grok Imagine agent mode beta!β
βand @townaiβ
βBe sure to check out the list. It is on our website right now.β
βTry Granite 4.1 here π Speech: https://replicate.com/ibm-granite/granite-speech-4.1-2bβ
βTry it here πβ
βThen, take any generated bipedal 3D character and automatically rig it in under 30 seconds with create-character.β
βwhat it is um and why you think it's interesting to look at yeah 100% Let me let me get my screen share pulled up and then we'll all right so we're we're in the GPT store um so I was lucky enough to bβ¦β
β3D animation and rigging models from @Uthana_Inc are on Replicate!β
βTry text-to-motion v1 and v2: https://replicate.com/uthana/text-to-motion-vqvae-v1β
βTry text-to-motion v1 and v2: https://replicate.com/uthana/text-to-motion-diffusion-v2β
βhe's the creator of Karass which is an open source deep learning library that is designed to enable fast user friendly experimentation with deep neural networks it serves as an interface to several deβ¦β
βBenchmark tests conducted on spiking neural networks, characterized by biological parameters, have demonstrated significant enhancements in both problem size and simulation performance, surpassing theβ¦β
βLike Llama 3 70B is a pretty good model.β
βCouncil Mode for comparing equity research put out by analysts from different investment banks, now supported on Perplexity Computerβ
βOur magic quantifiers may provide new tools for addressing some specific problems in magic resource theory.β
βcodex is also a better slack ai search than slack ai search.β
βTry out Grok STT here: https://replicate.com/xai/grok-speech-to-textβ
βTry Granite 4.1 here πβ
βTry out Grok TTS here: https://replicate.com/xai/grok-text-to-speechβ
βIn this work, we present SnapLog, an approach to extract event data from videos by converting frames to feature vectorsβ
βWe introduce HubRouter, a pluggable module that replaces O(n^2) attention layers with O(nM) hub-mediated routingβ
βWe present failure cases of symbolic evaluation in two popular frameworks, Lighteval and SimpleRL, and compare them to our approach, demonstrating clear improvements over commonly used methods.β
βWe present failure cases of symbolic evaluation in two popular frameworks, Lighteval and SimpleRL, and compare them to our approach, demonstrating clear improvements over commonly used methods.β
βTHIS REPOSITORY IS DEPRECATED. USE THE MODULE `keras.applications` INSTEAD.β
βThis repository contains code for the following Keras models: - VGG16β
βThis repository contains code for the following Keras models: - VGG19β
βThis repository contains code for the following Keras models: - ResNet50β
βOur core technical contribution is the \textbf{DDPO} algorithm,Diversity Driven Policy Optimization, a multi-turn GRPO-based approach designed to preserve dialogue diversity while holistically optimizβ¦β
βThis repository contains code for the following Keras models: - CRNN for music taggingβ
βyou can use flux or mlj.jl both of those are i think viableβ
βWe build upon this promising foundation and extend the method to work as an uncertainty estimation technique for already-trained artificial neural networks in the domain of regression. Our experimentsβ¦β
βTo learn more how this works, check out the Tailwind documentation [here](https://tailwindcss.com/docs).β
βBased on this observation, we propose QuantClaw, a plug-and-play precision routing plugin that dynamically assigns precision according to task characteristics.β
βWe validated our method, QDTraj, by generating diverse trajectories in simulation and deploying them in the real world. QDTraj generates at least 5 times more diverse trajectories for both hinge and sβ¦β
βOverall, H-optimus-1 achieves the strongest survival prediction performance.β
βNotably, the compact distilled model H0-mini slightly outperforms its larger teacher model H-optimus-0, despite using fewer than 8% of the parameters and enabling significantly faster feature extractiβ¦β
βThe platform is publicly available at Energy-Arena.org.β
βHDET is implemented as a drop-in replacement for PyTorch's OneCycleLR scheduler, requiring no changes to model architecture, optimizer, or data pipeline.β
βThe system uses instruction-tuned qwen3-embedding-0.6B embeddings...Qwen3 embeddings with 300-token chunk size achieved 94.6% accuracy on a clinical question-answering benchmark.β
βIts closure leaves behind non-updatable benchmarks, irreproducible results, and ultimately a field at risk of perpetuating these issues by turning to closed-source LLMs.β
βOxyGent is publicly available at https://oxygent.jd.com/β
βThe resulting corpus comprises Β± 35 billion tokens across the medical domain in about 100 million documents, freely available on Hugging Face.β
βthe most widely adopted multilingual base (Chatterbox, 23 languages) does not even tokenise Telugu or Tamilβ
βwe have a gpt-3 fine-tuning api these days um you know we'll be rolling that out for codexβ
βthe best open-source bases (Chatterbox, Indic Parler-TTS, IndicF5) trail them on measured phonological dimensionsβ
βFor intra-sentential code-mix we add a third branch (IndicF5 + native-script transliteration) that drops code-mix LLM-WER from 0.80-0.85 to 0.14-0.27β
βWe release R6 LoRA weights (Apache-2.0), inference code and router (MIT), and a Gradio demo.β
βSo there's one approach that I was very excited about and that I thought was it's very cool and I really like it's it's called dreamcoder by Dr. Kevin Ellis and and folks Um, so check it out if you ifβ¦β
βIn this paper, we introduce AegirJAX, a fully differentiable hydrodynamic solver based on the depth-integrated, non-hydro hydrostatic shallow-water equations.β