811
Reverse engineering a $1B Legal AI tool exposed 100k+ confidential files
Update: This post received a large amount of attention on Hacker News — see the discussion thread.
New
The latest blog posts to take off on Hacker News. 3,091 posts.
812
Are we repeating the telecoms crash with AI datacenters?
Looking at actual token demand growth, infrastructure utilization, and capacity constraints - the economics don't match the 2000s playbook like people assume
813
AI Is Breaking the Moral Foundation of Modern Society
I’ve been thinking about why AI feels like such a destabilizing force for society, and it isn’t just the threat of even greater…
814
LLM from scratch, part 28 – training a base model from scratch on an RTX 3090
I felt like it should be possible to train a GPT-2 small level model on my own hardware using modern tools and open datasets from scratch. It was!
815
Sycophancy is the first LLM "dark pattern"
People have been making fun of OpenAI models for being overly sycophantic for months now. I even wrote a post advising users to pretend that their work was…
816
A new AI winter is coming?
Though LLMs had a lot of promise, this has not been demonstrated in practice. The technology is essentially a failure, and a new AI winter is coming.
817
Google, Nvidia, and OpenAI
OpenAI and Nvidia are both under threat from Google; I like OpenAI’s chances best, but they need an advertising model to beat Google as an Aggregator.
818
Search tool that only returns content created before ChatGPT's public release
https://slopevader.com/ A search tool and browser extension for avoiding AI slop. Available for Chrome or Firefox. The tool filters web...
819
Don't push AI down our throats
AI is being done wrong.
820
Hachi: An Image Search Engine
For quite some time now, i have been working on and off on a fully self-hosted search engine , in hope to make it easier to search across Personal data in an end to end manner. Even as individuals, we are hoarding and generating more and more data with no end in sight. Such "personal" data is being stored from local…
821
The unexpected effectiveness of one-shot decompilation with Claude
A deep dive into how a 'one-shot' Claude headless loop (paired with scoring, defensive tooling, and a simple bash driver) massively accelerates matching decompilation of Snowboard Kids 2 on the Nintendo 64.
822
So you wanna build a local RAG?
How to build a RAG system that's fully self-hosted.
823
A trillion dollars (potentially) wasted on gen-AI
The machine learning community is finally waking up to the madness, but the detour of the last few years has been costly.
824
A Remarkable Assertion from A16Z
A friend made me aware of a reading list from A16Z containing recommendations for books, weighted towards science fiction since that’s mostly what people there read.
825
We're losing our voice to LLMs
We are losing something precious in the age of LLMs: our unique voices.
826
TPUs vs. GPUs and why Google is positioned to win AI race in the long term
I am publishing a comprehensive deep dive, not just a technical overview, but also strategic and financial coverage of the Google TPU.
827
The current state of the theory that GPL propagates to AI models
When GitHub Copilot was launched in 2021, the fact that its training data included a vast amount of Open Source code publicly available on GitHub attracted significant attention, sparking lively de…
828
The inefficiency of RL, and implications for RLVR progress
And implications for RLVR progress
829
I don't care how well your "AI" works
The other day I was sitting on the doorstep of a hackerspace, eating a falafel sandwich while listening to the conversation inside. The topic shifted to the use of “AI” for everyday tasks, people casually started elaborating on how they use “chat assistants” to let them write pieces of code or annoying emails. The…
830
The Generative Burrito Test
A critical benchmark for image generation models: A partially eaten burrito with cheese, sour cream, guacamole, lettuce, salsa, pinto beans, and chicken.
831
Ilya Sutskever: We're moving from the age of scaling to the age of research
“These models somehow just generalize dramatically worse than people. It's a very fundamental thing.”
832
The Bitter Lesson of LLM Extensions
From ChatGPT Plugins to Agent Skills, a look at how we've been trying (and failing) to extend LLMs for the last three years.
833
Three Years from GPT-3 to Gemini 3
From chatbots to agents
834
Agent design is still hard
My Agent abstractions keep breaking somewhere I don’t expect.
835
836
FAWK: LLMs can write a language interpreter
After reading the book The AWK Programming Language (recommended!), I was planning to try AWK out on this year’s Advent of Code. Having some time off from work this week, I tried to implement one of the problems in it to get some practice, set up my tooling, see how hard AWK would be, and… I found I’m FP-pilled.
837
Measuring the impact of AI scams on the elderly
We worked with Reuters to measure the impacts of AI scams on the elderly
838
AI is a front for consolidation of resources and power
Best case: we're in a bubble. Worst case: the people profiting most know exactly what they're doing.
839
Trying out Gemini 3 Pro with audio transcription and a new pelican benchmark
Google released Gemini 3 Pro today. Here’s the announcement from Sundar Pichai, Demis Hassabis, and Koray Kavukcuoglu, their developer blog announcement from Logan Kilpatrick, the Gemini 3 Pro Model Card, …
840
I caught Google Gemini using my data and then covering it up
I asked Google Gemini a pretty basic developer question. The answer was unremarkable, apart from it mentioning in conclusion that it knows I previously used ...