GM Frens

GLM 5.2 ranks #3 on FrontierSWE. It is only behind Fable 5 and Opus 4.8, and it outperforms GPT-5.5. This is the first model that closes the large gap between models from Anthropic / OpenAI and other providers, and it is the strongest open-weight model by far.

Today we reduced the size of the Matter Labs team. This was my decision, and I want to explain it. In 2024 we began building for regulated financial institutions. That work became Prividium, and the entire company is now committed to one goal: building the infrastructure that brings enterprises and regulated financial institutions onchain, with privacy at its core. As that work has progressed, we have learned a great deal about where our customers need it to go. Meeting that direction calls for a different mix of skills and roles than the phase before it, and some of what made sense earlier is not what we need now. That is the reason for today's changes. It hurts, because this is not about the effort or talent of the people leaving, who I deeply respect. They are some of the strongest engineers, designers, and operators I have worked with. Everyone leaving has been offered financial support, and we are helping with the transition wherever we can. If you are hiring, you can request access to our opt-in talent list here: docs.google.com/forms/d/e/1F… To everyone leaving: thank you for what you built here, and for the standard you set. Alex

We registered the AI agent itself with the SEC as an investment advisor. It has your complete context on your portfolio and account history. Speak to it in plain English to take action on your account. It will even prompt you with ideas you hadn’t thought of.

Zai was gracious enough to give me a key to test out GLM 5.2. I used it on a few simple tasks and quickly realized this model is on another level. I committed to using GLM 5.2 solely for the weekend and yesterday on everything from simple data analysis, random queries, side projects, and real work, and I can honestly say this is the first open model that I could comfortably replace Opus 4.8/GPT 5.5 with. It’s THAT good. When I say everything, I mean everything. I never needed to fallback to GPT 5.5 or Opus 4.8. This really blew my mind. I was unable to find any task where I knew GPT 5.5 or Opus 4.8 could solve, but GLM 5.2 could not, and I actually found a few cases where GLM 5.2 was better. I am not trying to overhype anything here. It's just my actual experience with this model. It was of course only 3ish days of usage, maybe cracks would form in time, but the perf is staggering imo. I see it's an "inferior" model on the benchmarks even Zai has shared, but I am not so sure and I think this is the first time I've experienced that with an open model. I am not saying it's necessarily better, but I believe it's a replacement that you could run on-prem, which is crazy to me. It was to the point where I was double triple quadruple checking that I wasn't accidentally running Opus or GPT. I ran thru both Hermes and my own custom coding agent harness with extremely great success. I cannot believe this is only a 754B model that's also an open MIT licensed model. Do not sleep on this one, and definitely try it out. Get it locally if you can! Can I find a way to run it locally? That’s a different question, but I will be trying to get it done because this model is epic.

GLM 5.2 is now live on Venice. @Zai_org's frontier open-source model built to power agents and long-horizon agentic coding, with High and Max reasoning modes for the hardest problems. Available fully private to Pro users.

Introducing GLM-5.2: Frontier Intelligence, Open Weights - Significant improvements in coding and agentic tasks - Strong long-horizon capabilities with a 1M context window - Two levels of reasoning effort: GLM-5.2 (max) pushes the limits, while GLM-5.2 (high) strikes a strong balance between performance and token efficiency - MIT-licensed open weights - Same API pricing as GLM-5.1 Tech Blog: z.ai/blog/glm-5.2 Weights: huggingface.co/zai-org/GLM-5… API: docs.z.ai/guides/llm/glm-5.2 Coding Plan: z.ai/subscribe Chat: chat.z.ai

hilarious setup honestly

BREAKING: SpaceX, $SPCX, officially surpasses Microsoft to become the 4th largest public company in the world, now worth $2.95 trillion.

[ ZOOMER ] BINANCE SET TO LOSE PERMISSION TO OFFER SERVICES TO EU CLIENTS: RTRS

The first real, 1:1 backed tokenized stocks are coming. → Own actual tokenized shares of U.S. companies → Trade, hold, and redeem - all onchain → Automatically receive dividends No derivatives, no IOUs. Welcome to the future of stocks.

In partnership with @stripe, Hermes Agent now supports a full suite of Stripe skills. Your agent can buy things, pay per-call APIs, and provision its own SaaS, with configurable safety limits on every action.

We're excited to join forces with @SpaceX to advance the frontier of useful AI. Expect significant improvements to Cursor soon.

$GEOD is probably the token I own that I've done the least modeling work on. Investment case is simply: Many robot coming. Robot need RTK. @GEODNET best RTK (and hard to replicate)

GM Frens

Robinhood cuts 10% of staff

SpaceX hit $3 trillion market cap today. This means Elon Musk made more money in the last 24 hours than Warren Buffett made in his entire lifetime. Insane.