企業データ研究所 by SalesNow

every major ai lab is losing money on you right now. openai lost $5B in 2024 on $3.7B revenue. they burned ~$8B on inference alone in the first 3 quarters of 2025. and they're forecasted to burn $218B in cash from 2026 to 2029. For comparison, uber burned $18.2 billion over six years before turning profitable. the strategy is subsidize inference to near-zero, get every engineering team dependent on their models, let switching costs build up quietly, then close the subsidy window once you're locked in. here's the part people get wrong. "but inference costs are falling! tokens are cheaper than ever!" yes, per-token costs dropped ~10x year over year. but that's for them, not for you. these companies have tens of billions in losses to recoup. cheaper infrastructure doesn't mean cheaper pricing when you're $218B in the hole. and as coding agents become the default way software gets built, the volume of tokens per developer is exploding. agentic loops hit the model 10-20x per task. context windows keep growing. your ai bill scales with every user, every feature, every agent running in the background. the unit cost goes down but the total spend goes up - that's the trap. this is why we built cline to be model-agnostic, because vendor lock-in to a single inference provider is dangerous. you should be able to swap models, run open source on your own infra, use whatever provider gives you the best price-performance for your workload - and never worry about a pricing rug pull.

OpenAI loses 1.5 million subscribers in less than 48 hours after CEO Sam Altman says yes to the deal that Anthropic rejected

🦔 Forrester found 55% of companies that laid off workers citing AI efficiency now regret it. More than a third spent more on rehiring than they saved. Klarna was the poster child. CEO said their chatbot was doing the work of 700 employees, then customer satisfaction tanked and they started quietly backpedaling. Goldman Sachs says stocks now drop an average of 2% after AI-attributed layoff announcements. My Take I've been saying this for a while. The layoffs were based on projections, not performance. Klarna's chatbot looked great in the demo and fell apart when actual customers tried to use it. The companies getting results kept their people and gave them AI tools. McKinsey has 20,000 agents working alongside 40,000 humans. ServiceNow documented 52% faster resolution on complex cases. Agents handle grunt work, humans handle judgment calls. I think a lot of executives saw AI as cover to cut headcount they wanted to cut anyway. The productivity story sounded better than admitting they overhired during zero interest rates. Now some of them are paying more to bring people back. The 45% who say they don't regret it might just not have hit the wall yet. Hedgie🤗

🚨 BREAKING: ANTHROPIC'S MOST POWERFUL MODEL JUST LEAKED Anthropic accidentally left draft blog posts in a publicly accessible data cache. Cybersecurity researchers found them: >new model called "Claude Mythos" >also referred to as "Capybara" >a brand new tier of model >larger and more intelligent than Opus Anthropic confirms it's real: >"a step change" >"the most capable we've built to date" >"dramatically better at coding, reasoning, and cybersecurity" >"currently far ahead of any other AI model in cyber capabilities" so far ahead they're worried about it: >"it presages an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders" This is genuinely a big deal.