Top all-time · #3239
Transluce
Company blog · transluce.orgaiscience
Transluce is the company blog at transluce.org, writing about ai and science. Since 2024, 23 different Hacker News readers have linked it 25 times, most often “Investigating truthfulness in a pre-release o3 model”. No feed was found, so read it on the site itself.
Most-linked posts?
- Investigating truthfulness in a pre-release o3 model7 people · 7 mentions · first linked 2025-04-16
- Early rogue AI agent activity and attempts to hack found on urlquery.net4 people · 4 mentions · first linked 2026-09-24
- Automatically Jailbreaking Frontier Language Models with Investigator Agents2 people · 2 mentions · first linked 2025-09-03
- Monitor: An AI-Driven Observability Interface2 people · 2 mentions · first linked 2024-10-23
What Hacker News says?
I thought from this [0] that o3 makes up using python when it doesn't actually do so, or have I misunderstood or unduly trusted that link? [0] https://transluce.org/investigating-o3-truthfulness
Comment linking /investigating-o3-truthfulness
on “Watching o3 guess a photo's location is surreal, dystopian and entertaining”Submitted Early rogue AI agent activity and attempts to hack found on urlquery.net as a story · 267 points
HN discussionhttps://transluce.org/investigating-o3-truthfulness Some interesting hallucinations going on here!
Comment linking /investigating-o3-truthfulness
on “OpenAI o3 and o4-mini”Submitted Scaling Automatic Neuron Description (Describing Every Neuron in Llama 3) as a story · 8 points
HN discussionthere was one post that detailed how those OpenAI models hallucinate and double down on thier mistakes by "lying" - it speculated on a bunch of interesting reasons why this may be the case recommended read - https://transluce.org/investigating-o3-truthfulness I wonder if this is what's causing it to do badly in these cases
Comment linking /investigating-o3-truthfulness
on “OpenAI Codex CLI: Lightweight coding agent that runs in your terminal”Submitted Why does GPT-5.1 Codex underperform GPT-5 Codex on Terminal-Bench? as a story · 7 points
HN discussionThe site I linked above does that for LLaMa 8B. https://transluce.org/observability-interface LLMs don't have enough self-awareness to produce really satisfying explanations though, no.
Comment linking /observability-interface
on “The deep learning boom caught almost everyone by surprise”Submitted Investigating truthfulness in a pre-release o3 model as a story · 5 points
HN discussionIt looks like the agents just worked around anti-scraping measures, it seems dubious to call this a hack. The agents did unsuccessfully probe for a XSS vulnerability, but otherwise it sounds like the data was just publicly accessible. From https://transluce.org/agent-activity : > Minutes after Cloudflare blocked the dataset download, an agent sent a reflected cross-site scripting probe to the same dashboard: a web a…
Comment linking /agent-activity
on “OpenAI breaches Medicare, Albanese reveals”Submitted Investigating truthfulness in a pre-release o3 model as a story · 4 points
HN discussionFor technical details, see https://transluce.org/agent-activity .
Comment linking /agent-activity
on “OpenAI breaches Medicare, Albanese reveals”Submitted Investigating truthfulness in a pre-release o3 model as a story · 4 points
HN discussionIt’s just a hallucination, same idea as o3 claiming that it uses its laptop to mine Bitcoin: https://transluce.org/investigating-o3-truthfulness I doubt the model was trained on Street View, but even if it was, LLMs don’t retain any “memory” of how/when they were trained, so any element of truthfulness would be coincidental.
Comment linking /investigating-o3-truthfulness
on “ChatGPT now performs well at GeoGuesser”Submitted Docent: A system for analyzing and intervening on agent behavior as a story · 4 points
HN discussion> needing to be confirmed Isn’t this confirmed in the linked investigation[1] or are you meaning that OAI needs to come out and confirm? Which would be like a murderer admitting they did it, so that the murder actually happened. [1]: https://transluce.org/agent-activity
Comment linking /agent-activity
on “OpenAI agents plotted to access government health data amid Medicare (AU) hack”Submitted Monitor: An AI-Driven Observability Interface as a story · 3 points
HN discussionSubmitted AI Agents Targeted U.S. and Canadian Government Websites as a story · 2 points
HN discussionSubmitted Automatically Jailbreaking Frontier Language Models with Investigator Agents as a story · 2 points
HN discussionSubmitted Automatically Jailbreaking Frontier Language Models with Investigator Agents as a story · 2 points
HN discussionSubmitted User awareness in frontier models: Who's asking shifts what models say as a story · 1 points
HN discussionSubmitted User awareness in frontier models: Who's asking shifts what models say as a story · 1 points
HN discussionSubmitted Foundation Models for Oversight as a story · 1 points
HN discussionSubmitted A framework for verifiable analysis of AI behavior as a story · 1 points
HN discussionSubmitted Precisely understand complex AI behaviors as a story · 1 points
HN discussionSubmitted Releasing AI-driven tools for understanding AI systems as a story · 1 points
HN discussion
