Tag
The video criticizes Anthropic executives' claim that "coding is solved," pointing out that Claude Code has basic bugs like terminal flickering, mysterious error messages, and session leaks that have persisted for over a year, arguing that Anthropic is either lying or deluding itself.
An observation that excessive carefulness is the most predictive trait for successful research, but AI agents often make it harder to find bugs.
The article draws parallels between the security flaws in C/C++ and those in Verilog, arguing that the hardware description language's design leads to bugs and that the industry should invest in safer alternatives, similar to the push for memory-safe programming languages in software.
Boris Cherny announced that Claude Code now defaults to self-prompting, but a critic notes the settings.json has over 125 keys, mostly undocumented, and includes bugs like the deny-rule bug and a 10% Opus tax.
This article details over 20 security vulnerabilities found by AI agents in Epsilon, a small WASM runtime written in Go, including several sandbox escapes that allow malicious modules to break out of isolation.
Bjarne Stroustrup criticizes AI-generated code, stating it introduces more bugs, bloat, and security holes, and is nearly impossible to validate, with senior developers retiring rather than dealing with it.
A blog post recounting a bug where inserting adjacent multi-byte emoji caused a splice in a CRDT library, splitting a surrogate pair and silently breaking collaborative editor synchronization.
Luca Ronin launched Tolaria yesterday, receiving 5K+ social media likes/bookmarks and 800 GitHub stars, along with many bug reports.