@DarioCpx They've exhausted all the free data scraped from the historical internet. The new inputs are becoming recursive like a photocopy of a photocopy.

Something has gone badly wrong in both Anthropic and OpenAI in the past months despite both companies raised a biblical amount of capital

@DarioCpx Compute/cost? It seems they’re pulling more cached answers than ever before, and it’s using old data. Asked Claude Opus about the price of a stock and it was a full day behind the actual market value. It then proceeded to double down on being wrong.

@DarioCpx Yes Opus 4.8 is way worse than 4.6 I dont know what is really happening 🤷🏻‍♀️

@DarioCpx Did anyone ask for "ai" to be implemented into every SaaS Fn app? No. Does it deliver value? No. They did it anyways. Now they complain it doesn't work. Good luck.

@DarioCpx people behind the curve still think AI will replace us all.. well, let's hope it will at least replace them 😂

@hidflect @DarioCpx Recursive data is basically model drift with better branding.

@DarioCpx Because you can’t automate quality control and craft as much as they wish they could

@DarioCpx If you thought the dot‑com IPO wave was a liquidity extraction event dressed as a revolution, wait until you see the 2026 version: three trillion‑dollar money pits going public in the same quarter, backed by Bermuda SPVs, GPU‑backed annuities,

@DarioCpx They need to spend more tokens on whistle blower culling.

@DarioCpx Hey, don't forget that Gemini has gone to shit too!

@DarioCpx Confirm! We are now trying to use others as default llms. Hope solution will come as this far they were best choice.

@DarioCpx Circular slop reference

@DarioCpx Higher numbered Models burns much more tokens too.

@DarioCpx True in day to day work

Gpt 5.5 went completely retarded where opus 4.8 is little bit good

@DarioCpx @AnthropicAI is going to be the greatest short of all time. If you look under the hood, the generated code of Opus 4.7 and 4.8 are truly slop.

@Noahpinion Yes. Parallelizable agentic workflows with 1m token contexts are compute intensive/enterprise adoption has picked up tremendously in the past 6 months. There isn’t enough infra to comfortably support it - thus the $200b Anthropic committed for TPUs + xAI data center access.

@DarioCpx Are we finding out AI is just SaaS

@DarioCpx There are training their next models on customer platform.

@DarioCpx I can’t even get ChatGPT to get the timing on cooking a turkey right