Post #2798939

@dolmen@mamot.fr @tagomago@mastodon.social @davidgerard@circumstances.run @stefano@mastodon.bsd.cafe I think the real world use case for agents is going to be a handful of small projects that get billed a few thousand dollars by the LLM provider (Anthropic, ChatGPT etc) to make something that's riddled with security vulnerabilities and other major issues. We've watched the agents just haphazardly adjust tests to bypass major testing criteria as is just to get the test to pass, and I only foresee that type of behavior continuing unmitigated. Agents are the current phase of Hype for genAI and they're going to be slightly better at pretending to do work, but until I see open source commits coming it from Agents, I don't believe a single thing about that's being promised.