Peter Sergeant

Those drugs are the best thing to happen to me

Forget the complexity: AI all boils down to drawing the right lines | sgnt.ai Over and over again, despite the best efforts of humans, the most effective AI systems come down to one simple idea: finding the right shaped line that fits some data points.

Forget the complexity: AI all boils down to drawing the right lines: sgnt.ai/p/finding-li...

ChatGPT gave me what as far as I can tell is a genuinely novel idea for an algorithm (well, a novel application of a well-known algorithm) and now I’m having self doubts

I saw some guy saying they refuse to answer about “crime statistics” and chastise him for asking, which was a head-scratcher

I think that’s prima facie not a workable approach in practice, but, criminal liability for directors and fines that make bond owners take a huge loss could work

The worst part is the dramatic medication changes often happen when you’re already not at your strongest so it can be hard to pick apart 🤷‍♂️

I ran a small boutique recruitment agency as a side project for a few years and it was highly fulfilling (people are so grateful on both sides, improving people’s lives), very lucrative (£100k a year), and really didn’t take that much work (1-2 hours a day). If you have the network, do it!

Mapping US troops and military bases in the Middle East There are between 40,000 and 50,000 US troops in the Middle East across at least 19 sites.

The region is absolutely stuffed with US bases and detachments … with the notable exception of Israel www.aljazeera.com/amp/news/202...

I mean it’s theoretical because consumers can (usually) shop prices, whatever Delta’s exec fantasies might be

Understanding Modern AI is Understanding Embeddings: A Guide for Non-Programmers (with lots of dogs!) | sgnt.ai Embeddings are a core AI concept that underpin a great deal of what we today think of as being AI. This article is going to give you an accurate and intuitive understanding of what an “embedding” is i...

Spread the joy! sgnt.ai/p/embeddings...

Proof of concept that you can get an LLM to disbelieve its own lying eyes

Well, and the media

Exclusively screenshots on Bluesky 😅

Technically this is pretty interesting, because nobody has as much high-quality training data as they want, and so nixing high-quality reporting as an input source and replacing it with Stormfront is already going to be a challenge…

In RAG, your chunks solve three distinct problems; trying to use one algorithm to solve all three probably isn't optimal: sgnt.ai/p/rag-trinity/

I think part of the problem is that it’s possible to curate your feed pretty well to get rid of the hate-bait and just see what you want. It’s not until you try going back as a preference-naïve account that you see how bad it’s become

RAG chunking isn’t one problem, it’s three | sgnt.ai Existing articles often focus on a chunk as a singular concept: you split the article into paragraphs, say, and use these to feed the LLM, generate embeddings, and quote back to the user. But that's t...

I have an unfinished draft that mentions a similar technique here: sgnt.ai/p/rag-trinity

This is a lovely way of putting it and I agree

Process and accountability issue imo

GitHub - pjlsergeant/xslt-fever-dream: XSLT is a great programming language XSLT is a great programming language. Contribute to pjlsergeant/xslt-fever-dream development by creating an account on GitHub.

One day in 2012 I wrote up from a nightmare in which I was being chased by angle brackets, and sweating profusely, and found this code in my editor github.com/pjlsergeant/...

I will fix your vibe-coded MVP | sgnt.ai You delivered your vision quickly using one of the many excellent LLM-based coding tools, users are happy. I will dig you out of the tech-debt hole you have created.

I will fix your vibe-coded MVP. Not joking. sgnt.ai/p/vibe-coded/

I like the LaTeX explicit double hyphen as an alternative --

Can we start calling people “AI deniers”?

Oddly I find ChatGPT too combative, and Claude to be too persuadable. Gemini answers 1 in 5 queries in the style of a press release. What an amazing time to be alive

“At least two of Iran's neighbors — both Islamic nations — already have nuclear weapons” … who’s country two here? Turkey?

Gender and Positional Biases in LLM-Based Hiring Decisions: Evidence from Comparative CV/Résumé Evaluations This study examines the behavior of Large Language Models (LLMs) when evaluating professional candidates based on their resumes or curricula vitae (CVs). In an experiment involving 22 leading LLMs, ea...

I think the most interesting paper I've seen has been the favouring of women's CVs over those of men by LLMs, which I suspect is the result of a massive pro-man bias in the training data being squelched by somewhat blunt RLHF; arxiv.org/abs/2505.17049

Cambridge (UK) has some of the country's best Chinese food and attracts some of the brightest minds because Xu Zhimo liked watching punts go by, and the Trump admin wants to screw up the Cambridge (USA) version of this

Harvard Has Trained So Many Chinese Communist Officials, They Call It Their ‘Party School’ The university’s Kennedy School of Government has long been favored by party cadres seeking career boosts.

China sends its most promising civil servants to a US university for 1-3 years, and rather seeing this as a total and complete Cultural Victory, recruitment opportunity, and chance to indoctrinate them, the Trump administration wants to shut this down. Baffling. www.wsj.com/world/china/...

Being one of a small number of remote workers for an otherwise in-person team can be very tough. I would have to think very carefully before joining another team like that!

A real shame, you’re one of the accounts I enjoy following the most