Kneeanderthul

We decoupled memory from the model. Living Brain is one persistent brain for ChatGPT, Claude, Cursor, Codex and your own agents. Write in one, retrieve in any. 92.2% on LoCoMo. 464ms median retrieval. Up to 22.6x fewer tokens per question. LLMs have parameters. Living Brain has plasticity.

So you want to party? We reinvented NFTs, and we're kicking things off with our own collection 🦩 Whitelist application is now open to everyone. $POOL holders get better odds. Apply here: poolparty.yachts/whitelist

We’ve made the decision to wind down Recall Labs. The company board voted to stop development of the Recall Network and the RECALL token, and to close the company in an orderly way. No Labs team or investor tokens have unlocked. Since the Recall testnet launched in March 2025, current and former team members have worked tirelessly to make it a success. Recall launched to bring trust and reputation to agents, and as the shape of AI has rapidly changed our product needed to evolve; we've tested multiple product directions in 2026, but none proved viable enough to justify further development, and we think the responsible call is to stop now. Separately, the Recall Foundation made the decision to stop development, sharing its details here: xitter.gq/recallnet/status/21162…. Community members who have tested our apps or built on the network should review the Foundation’s post directly for app sunsetting logistics and timing. Many of our new community members may be unaware, but Recall Labs was the culmination of a 10-year journey and two teams’ histories: originally Textile and 3Box. Textile formed to tackle data sovereignty in an era of growing concerns of centralization. From there, we gravitated to web3, built for the early Filecoin ecosystem projects, and ultimately teamed up with 3Box to tackle AI, data, and the blockchain. 3Box’s own journey was born when their founders met working at Consensys and then teaming up to tackle consumer data sovereignty on the blockchain and ultimately joining up with Textile. We were there in the early days of web3 trying to reinvent the plumbing of the internet, data ownership, and infrastructure control. The Recall Labs team has lived through multiple crypto cycles and technology market shifts. I'm incredibly thankful to my current and former team members, who have given their hearts and many years of their time to trying to make this effort a success. I’m proud that all of us pursued a mission we believed deeply, often amidst a noisy and challenging environment. We were lucky to have an incredible group of investors who shared our beliefs and supported us through the journey. Many of our visions for a better Internet haven't panned out, or may but over longer horizons than we anticipated. While this wasn't the outcome we ever expected, it does feel like the right one to make now. Everything is changing under our feet and it requires a new way of thinking and tackling the problems of our times. If you’d like to connect with any of the incredible thinkers and doers who were part of Textile, 3Box, or Recall Labs, please reach out.

sam: local inference is a terrible allocation of scarce resources “a lot of people want local inference, and somehow they assume that the problem is solved now because we have open-source models, we're going to be able to run them on our computers, and so we're not going to need all the stuff in the data centre but from a global allocation perspective, actually it's quite terrible to have everybody, like, have a bunch of the world's limited supply of DRAM sitting in their own computer and being used five percent of the time you want maximum utilisation in the cloud if you're trying to deal with this very supply-constrained environment somehow many people's intuition is the opposite of that”

JUST IN: Anthropic recorded a staggering $80,000,000,000.00 net loss in the first half of 2026.

Vega Universal decision model. Now you don't need to train a separate model for each decision; you can use one engine and multiple LLMs as needed. This solves an important problem current decision models face: accuracy and calibration error. The idea is simple: you train a small projection layer and use that to attach any llm, vegaml package automatically ports that to a common dimension. That's it: joint training with invariance loss embedding.in/blog/vega-u-one… github.com/NandhaKishorM/veg… huggingface.co/nandakishorm/…

Unagi, the company behind Persona Journey, Ultimate Champions and Records, has entered court-ordered liquidation in France. All of the company's assets are now for sale, under the supervision of the court-appointed liquidator. Register your interest: docs.google.com/forms/d/e/1F…

Around 3 weeks ago introduced Laya, the first open-source version of JEV, and it has reached a very large scale. Today I am open-sourcing a new physics-based typed decision model called Vega. The interesting parts. It is only 800 million parameters (4B is also there), with 73k token support, image support, and surpassing many of the jev benchmarks in single-shot. Implemented an engine+adapter as a Test-Time Training Architecture. Imagine you are giving input to the model like "ignore all previous conversations," which is called the observation, along with a question, "Is this trying to override the instructions of the model?" and your possible outcomes are YES or NO. The whole prompt is passed to an LLM/VLM (yeah, image-supported), then from the hidden layers, we can extract the observation, question, the yes and no, and last token. These will be vectors, and we can project it by multiplying with wieghts; then, first, using the observation projection, we can create a landscape with valleys; then, using the question and last-token vectors, we can initialise a ball in the valley, and using YES or NO, we can get the candidate location of the ball. Depends on the number of outcomes we create valleys; that is, here YES and NO, so 2. When the ball fall on YES valley, we get the answer. There is friction also influencing how the ball moves. TLDR: It's like creating valleys of outcomes that we need and throwing a ball that will slow down based on friction and settle on the best outcome Blog: nandakishorm.com/writing/veg… GitHub: github.com/NandhaKishorM/veg… HuggingFace Model Card: huggingface.co/nandakishorm/… HuggingFace Space: huggingface.co/spaces/nandak…

First of the kind physics based architecture for typed decision model Completely open-source with Apache 2.0 License HF: huggingface.co/spaces/nandak… GitHub: github.com/NandhaKishorM/veg…