Top 3 Themes from the Discussion
| # | Theme | Supporting Quote |
|---|---|---|
| 1 | Skepticism toward Anthropic’s “Mythos” security stance | “The whole point of Mythos/Glasswing is our best models are scary good at security research, so much so that we won’t let them help you find vulnerabilities unless you are a trusted partner.” — simonw |
| “Considering Anthropic had a sandbox‑bypass vulnerability in CC for a year, silently patched it, and still hasn't made a disclosure statement, no one on Earth should trust them or believe a word they say.” — nullbio | ||
| “Trying to work around Anthropic blocking security‑related prompts does get pretty tiring though.” — vachanmn123 | ||
| 2 | LLMs as tools for security research – promise and limits | “Nice writeup. A practical example of a project, what was found, how it was found, the quality of the findings, reproducible.” — tptacek |
| “The bad thing about software is that there’s infinite ways to solve the same problem… I’ve not had much success with them writing code that simply has no bugs.” — onlyrealcuzzo | ||
| “If you can get better at identifying complexity, LLMs can get much better at suggesting viable fixes.” — amelius | ||
| 3 | Community dynamics & incentives – authenticity vs. career‑building | “People want promotions, money and a job in general, and they will do stupid stuff to keep their jobs and increase their pay.” — hootz |
| “It is not at all, in the slightest, weird heuristic to deploy in the Agentic Era. It’s a heuristic after all. There is no proof one way or the other.” — tptacek | ||
| “Is there no room in your model of the world for someone to figure out something interesting using AI tools and then write about it just because they like sharing interesting information?” — simonw |
Summary:
The conversation clusters around (1) distrust of Anthropic’s restrictive security policy, (2) a realistic appraisal of how effective LLMs are at uncovering bugs, and (3) the broader HN culture where career incentives and authenticity are frequently questioned.