Wall Street just erased $2 TRILLION from software stocks and gave the wreck a name: The SaaSpocalypse. Our lead crypto analyst, @m0xt_, thinks the market just made a $2T sorting error, and he published a full report on where the money gets made (save this). Let's talk damage first... $CRM fell about 30%. $WDAY fell 33%. The main software index lost more than 20% in a single quarter. And for the first time on record, software trades at a lower earnings multiple than the average S&P 500 company. The market's golden child of two decades is priced like a dying industry. The accused killer is the AI agent: Software that does the work itself instead of helping a human do it - It resolves the ticket... It reconciles the invoice... So if an agent can do the job... Why pay for software that just helps a human do it? The bear case is serious, and @m0xt_ gives it real respect: 1. Microsoft CEO Satya Nadella said most business apps are basically databases with some rules on top. If an agent can talk to the database directly, the screens in between lose their reason to exist. 2. Per-seat pricing is built on a number that AI shrinks. Atlassian just logged its first ever decline in enterprise seat counts. Workday cut 8.5% of its own workforce. 3. Cursor went from zero to $2B in annual revenue in roughly 3 years, the fastest ramp in business software history. 4. The selloff's biggest leg down came when Anthropic launched Claude Cowork, an AI desktop worker that runs multi-step workflows on its own. But @m0xt_ found four cracks in the funeral story: Deutsche Bank says it doesn't know of a single software company expecting AI to hurt revenue this year. In Q1 2026, 14 of the 16 software names covered by one Morningstar analyst beat on both the top and bottom line. In the real software busts (2001, 2008, 2022), earnings collapsed with the stocks. This time the stocks fell and the earnings kept climbing. That gap has to close one way or the other. → The agents aren't ready just yet. Gartner predicts over 40% of agentic AI projects get canceled by the end of 2027, and says only about 130 of the thousands of vendors selling "AI agents" are the real thing. The rest are slapping the agent label on old chatbots. → The poster child of AI adoption changed its mind. Klarna cut hundreds of software tools and 1,200 employees after building an AI support system it said did the work of 700 people. Then quality slipped, Klarna started rehiring humans, and the CEO publicly said he doesn't think this is the end of Salesforce. → Nobody is actually leaving. No churn wave in Q1, and not one major vendor has reported customers walking away for home-built AI. So the market is pricing a takeover by a workforce of agents that, today, mostly can't be left alone with the keys. The one-sentence thesis from @m0xt_ goes like this: AI kills software that helps humans use tools, and feeds software that gets work done. The market is pricing both for the same funeral. Companies poured an estimated $30-40B into generative AI, and one widely cited MIT study found 95% of corporate AI pilots produced no measurable return. A model produces words and suggestions, but a business needs the invoice reconciled and the ticket closed, with a record of who approved it and rules about who was allowed to do it. Raw intelligence is cheap → trusted, finished work is not. And the selloff just put the companies that sell finished work on sale. Software trades around 23x forward earnings, down from over 80x at the 2021 peak. Goldman Sachs CEO David Solomon called the selloff "too broad." When everything gets sold for the same reason, the companies the reason doesn't apply to go on sale by accident. @m0xt_ is testing the thesis on three names: $NOW (ServiceNow) runs the plumbing of big companies: IT requests, HR cases, security incidents. A 98% renewal rate means customers almost never leave. Contracted future revenue hit $28.2B, up 26%. $CRM (Salesforce) is the stress test. If SaaS were truly dead, this should be the most exposed name on the board - and the market treated it that way: the stock fell roughly 50%. Its agent product, Agentforce, sits near $800M in ARR, up 169%. $TOST (Toast) runs restaurants - orders, payments, payroll. Roughly half its recurring revenue comes from payment processing, so it earns when its restaurants earn. You can't seat-compress a payments stream. Each name comes with exact tripwires in the report: The numbers that prove the thesis right, and the numbers that would make @m0xt_ admit he's wrong and cut. And he didn't stop at writing. Off the back of this report, he opened a brand new position in his portfolio: One of the three stocks above, at a specific entry price, with preset triggers for buying more and for selling. PRO members can see which stock, the entry, and every trigger right now. It costs $1 to find out which one he bought. Link in the first comment 👇

The Rio 3.5 model broke the internet this week. The plot twist? It’s essentially our open-source model, Nex N2 Pro, wearing a different hat. 🤯 We analyzed the weights, and the recipe is exact: Rio 3.5 ≈ 0.6 * Nex N2 Pro 0.4 * Qwen 3.5 It even literally introduces itself as "Nex N2 Pro" if you ask it without initial system prompt! 😂 We are flattered that the City of Rio used our work to achieve SOTA performance. Thanks for the ultimate benchmark validation. 🤝 But in the open-source world, attribution matters. 👇 Full mathematical proof & verify script in the first reply!

Alibaba Qwen3.7 slowly fading into irrelevance at the frontier due to proprietary stance. In it's place we have Minimax M3 and... *checks notes* Rio 3.5 397b, made by the municipal IT company of Rio de Janeiro's city government. huggingface.co/prefeitura-ri…

SITUATION DETECTED: The city of Rio de Janerio has post-trained a model. Based on Qwen 7/2, Rio 3.5 Open 397B adds SwiReasoning on top of the base Qwen model — a framework that dynamically switches between standard chain-of-thought and latent-space reasoning, guided by entropy-based confidence signals, so the model only "thinks out loud" when it needs to and otherwise reasons silently in hidden space for better token efficiency.