Going from crypto to fiat and back is an extremely monitored and regulated route. It might be an easy way to settle between counterparties, but a difficult one to launder.
This kind of company does not want "juniors who could independently own products end to end without handholding", they want do-it-all people on the cheap. They just don't know what it takes to do things the right way. It's not unusual that the organization is a mess as well, because management is unable to organize it.
Get out of it, this kind of companies have a dysfunctional management. After the initial learning, they will be unable to recognize your value and contribution.
Also, don't feel guilty about getting burnt out. It's not your fault and you got tougher.
Some may criticize regulations, but the EU-mandated cyber-resilience act (CRA) actually forced companies to have a clear contact point for vulnerabilities reporting, and to act upon it.
2026-09-11, save the date folks. That's when all companies selling products with digital elements in the EU have to have a reporting pipeline for actively exploited vulnerabilities and severe incidents.
I second this idea: LLMs will plateau. They are already pretty good. Plus, scientists struggle to actually score their performance accurately (esp. when it comes to reasoning).
With that said, they are now hitting the walls of energy costs and memory shortages. You brain uses 20W -- don't take it as an insult. There are orders of magnitude to gain from producing energy-efficient models (or model runners).
So I am expecting same performance at lower costs for the coming years.
AI has already replaced a lot of workers, but there is still no AI-managed business out there, regardless of doomsday PR.
I think that there is a misconception about what money is. It is a vector of value, and value comes from the work of the people. If less people can work, this will lead to deflation, something that capitalists would avoid. But remember that AI is hardware and energy, and that requires more workers. Your token price pays for electricity and hardware GPUs -- only marginally for AI science. Sure, developers have to be more like architects than code monkeys, but I am not sure if it is a bad thing.
Also, I have this contrarian view that the LLM tech will now plateau. They are not a path to AGI. Look at how they work, and you'll understand that they are unable to innovate. Models are like a compressed version of Internet knowledge, and that's what they are spitting out. That's already pretty good. But I don't think that we'll see another leapfrog innovation on LLMs anytime soon. After all, OpenAI happened by accident.
In the future, we will all have today's frontier AI on our laptop to automate our lives. There are still many things to invent out there, and I see LLMs more like an enabler than a competitor.
If your purpose is learning, use paper. Studies show that it much more efficient than screens to remember information, as your brain physically maps where the info comes from.
reply