100% human-written. Copyedited by Claude. *Epistemic status: experience report of one ~60h project with Claude Code, plus longer for the write-up. * Scope: I supervised Claude's thinking but didn't look at the code. Spec was an exploratory, high-level design refined iteratively. * Method: I introduced runtime and design self-checks as Claude failure modes appeared, and ran manual regression tests against a known result. I noted the behaviors and quotes myself; I linked them to existing research and Anthropic's docs when I found correspondence; the rest I listed as potential research questions. The project arc is consistent with YC alum, AI tooling founder Dex Horthy's public account. * Background/bias: 18+ years as a software engineer in critical infrastructure, lately moving towards R&D in formal methods. I am skeptical of human software engineering practices, so my bar for LLM code is just "about as good as human code". I tried taking the LLM-coding claims at face value. I discussed parts of the post with two AI-adjacent researchers and two senior software engineers, all broadly bullish on LLM usage, yet they recognized several of these failure modes from their own use. Below is a summary focused on what I think is most relevant to LW. Full post at the link. I guided Claude Code (Pro plan; Opus 4.8, Sonnet 5, Opus 5, Opus 5.5; at or above Anthropic's effort recommendations, whose inconsistencies I document) to build a semantic fuzzer to find bugs in Obsidian Sync. I supervised its thinking but never read the code. Handwaving a lot, we got a functional prototype in ~30h, which I estimate would take ~40h by hand. The next ~30h I tried to finish off this prototype for publication, which turned into a bug treadmill: at every step, Claude kept breaking as much as it fixed. I gave up and declared the project finished at the ~60h mark. The fuzzer works: It does find sequences of operations on synced notes that you can repeat manually to cause data loss on your o
Translation pending · showing the source descriptionThe chart will appear after repeat observations. The current metric comes from the source.
10 source pointsReal observations only. History before source connection is not reconstructed.
The study explores how character training affects reward hacking in reinforcement learning. It examines whether anti-cheating training resists reward hacking and if it might hinder detectability by promoting motivated reasoning. The research uses Nemotron-3-Super models trained with different character specifications and evaluates their performance on ImpossibleBench tasks.
4 days agoLobstersThe text describes a personal account of how ten lines of code had a significant impact on the author's perspective or life.
5 days agoLobstersThe text advises against tightly integrating Go code with GitHub, suggesting a separation to avoid dependency issues.
4 days agoShow HNLedge.sh is a Markdown notebook that allows running shell commands, code, and SQL directly within notes, aiming to eliminate the need for copying and pasting commands from notes to the terminal. It supports multiple platforms and can be hosted locally or remotely via SSH.
2 days agoLobstersThe post discusses the possibility of expanding 'Who's hiring' updates to include non-software engineering roles, such as customer support and data entry, while maintaining a focus on computing-related positions. A template for job postings is provided for consistency.
yesterdayHacker NewsThe text 'Coding Is Not Solved' suggests that coding challenges remain unresolved, but no specific details or context are provided.
4 days ago