HN user

etherio

882 karma

https://www.uzpg.me

halcyon (at) disroot dot org

Posts98
Comments115
View on HN
fulcrum.inc 13d ago

Fable is SOTA at CIFAR Speedrun: lessons on AI R&D automation

etherio
4pts0
fulcrum.inc 1mo ago

Agents are under-elicited: A case study in optimization tasks

etherio
2pts0
fulcrum.inc 1mo ago

Inverse Rubric Optimization: A testbed for agent science

etherio
30pts0
fulcrum.inc 2mo ago

Multi-agent systems as distributed software

etherio
4pts0
github.com 3mo ago

Show HN: Druids – Build your own software factory

etherio
64pts15
github.com 3mo ago

Show HN: Druids – coordinate and deploy coding agents across machines

etherio
9pts3
fulcrumresearch.ai 4mo ago

More Is Different for Intelligence

etherio
5pts0
orpheus.dev 5mo ago

Show HN: Orpheus – PR review that runs the code

etherio
1pts0
zephyyr.substack.com 6mo ago

2025 Letter

etherio
1pts0
www.uzpg.me 6mo ago

Personalization Requires Data

etherio
3pts0
fulcrumresearch.ai 7mo ago

Show HN: Lunette – auditing agents for evals and environments

etherio
1pts0
www.uzpg.me 8mo ago

Dense reconstruction is the scaffold of machine learning

etherio
1pts0
github.com 8mo ago

Show HN: Quibbler – A critic for your coding agent that learns what you want

etherio
114pts27
github.com 9mo ago

Show HN: Orchestra – an interface to usefully run coding agents in parallel

etherio
1pts0
fulcrumresearch.ai 9mo ago

Show HN: Automated monitoring and orchestration of coding agents

etherio
3pts0
fulcrumresearch.ai 10mo ago

AI Agents and Painted Facades

etherio
5pts0
zephyyr.substack.com 1y ago

The machine has no mouth and it must scream

etherio
1pts0
www.uzpg.me 1y ago

2024 in Review

etherio
1pts0
alexkritchevsky.com 4y ago

Aphantasia, or how not to do math in your head (2019)

etherio
3pts2
www.benkuhn.net 4y ago

My favorite essays of life advice (2020)

etherio
154pts10
www.uzpg.me 4y ago

Using AI and Web content to improve organization and discovery of my knowledge

etherio
2pts0
github.com 4y ago

Show HN: Espial – AI tool to automate discovery and organization of knowledge

etherio
19pts4
en.wikipedia.org 4y ago

Vesna Vulović

etherio
1pts0
www.uzpg.me 4y ago

2021 into 2022 – A Year in Review

etherio
1pts0
bewrong.substack.com 4y ago

The Expected Value of Friendships

etherio
1pts0
matt.might.net 4y ago

My Son's Killer

etherio
2pts0
en.wikipedia.org 4y ago

Livyatan

etherio
3pts1
en.wikipedia.org 4y ago

Cloud Iridescence

etherio
1pts0
en.wikipedia.org 4y ago

East German Balloon Escape

etherio
2pts0
en.wikipedia.org 4y ago

Rescue of Roger Mallinson and Roger Chapman

etherio
31pts1

Yeah, I guess we have sandboxes qwith our various code environemnts, and then we've been developing programs that run agents to do various things, which we iterated on to make them better at the task.

For example, one spawns copies of the env with a PR and runs agents in the dev env to verify by running and demonstrating functionality and then comments on github

another one is just a generic software factory that spawns a bunch of agents to coordinate on some repo, others do a redteaming flow, etc...

I'm researching how we can use technology to improve the way we learn and think.

I'm currently exploring how AI/Natural Language Processing can automatically organize personal knowledge and notes by tagging and linking ideas / content together [0], removing the friction of organization to help you find new connections and ideas.

If you're interested in this domain, I'd love to talk to you!

[0] https://github.com/Uzay-G/espial

I agree. I think there are several reasons high achievers' mental health issues go untreated:

- like you say they maintain a healthy surface because part of their perfectionism is no one knowing about their anxiety / having a good image

- if they do share, often their goals / expectations will be inflated compared to that of their peers such that for others it feels like they're bragging or being ridiculous, instead of taking their pain seriously. I notice this especially with when I share dissatisfaction about my school results - disparate expectations create a true divide.

Hey HN! A while back I posted about Archivy, a knowledge base I was working on that encouraged archival of online content (https://news.ycombinator.com/item?id=24199419).

I realized that one of the reasons knowledge management is overpowered is that you can create this tapestry of knowledge and ideas compounding off each other, and that's even more true if you integrate it with content from online.

This process is useful but it takes time to link and tag. So I decided to try and see how we could automate it!

Espial creates an auto-generated organization of our knowledge, conceptually and structurally.

This allows you to discover new connections in your ideas and search them based on meaning, not keywords.

The show hn links to the code, but I explain more of Espial's goals here: https://www.uzpg.me/tech/2022/01/31/redefining-pkm-with-nlp

Hope you like it!

I'm interested in note-taking / improving the way we learn for the following reasons:

- Having a trace of what I've learned that I can go back to. I consume loads of content (books, articles, etc...) and want to be able to use their insights in my future discoveries and general creative process. Being able to go through my ideas in my notes helps with this.

- that point links back to the idea of optimizing for discovery. I find that I can find much better ideas when I have a nuanced perspective and can oppose different content, so having an organized knowledge base allows me to link and discover new hierarchies in my thinking to enhance it. It's also why I'm building tools to automate connections between notes with AI, and it's why my knowledge base is heavily integrated with content I consume online, for example this post that I can now search in my knowledge.

- I think there's untapped value in sharing notes and tidbits of knowledge with other people. Blog posts take more time and have different aims than notes, but notes are also valuable / can be useful for others. My public wiki is at https://knowledge.uzpg.me

- having a time-logged track of what I've been learning: I like tracking my goals / progression (for eg yearly reviews https://www.uzpg.me/general/2022/01/01/a-year-in-review-v2.h...) and going through my knowledge base helps with this

Exhalation and Stories of your life and others by Ted Chaing. These collections of short scifi stories are stunning, and particularly the stories Exhalation and Story of your Life. They combine detailed technical descriptions of the science fiction aspects with raw emotional beauty. The movie Arrival is based on Story of your life.

La Peau de Chagrin (The Wild Ass's Skin) - Balzac. Beautifully written novel about desire and health, with verbose descriptions that I personally loved. Some nice quotes here [0]

Ubik by Philip K. Dick - Pretty mindbending and interesting twist on the whole "(distorted) perception of reality (what even is reality?)" theme. (9/10) [scifi]

[0] https://knowledge.uzpg.me/dataobj/2771/#quotes

I use https://github.com/lucasrla/remarks, which OCRs (text recognition) your highlights to extract what exactly was highlighted, and also outputs screenshots of all pages on which I wrote notes.

This way I can go through my annotations sequentially, save highlights / their main ideas, and reformulate my notes into plaintext a bit more clearly.

This is just subjective, but I use it to read many books and enjoy it. I can scribble / highlight / notetake very easily and find the whole experience quite convenient.

Where did you hear that?

I do this too using my remarkable (reader tablet )and then scripts that export highlights / notes which I then index for as plaintext - it makes this process much more seamless.

I think a lot of this is true not only for books, but also for articles.

I've started trying to be more selective with the online content I choose to spend my time on, and then conserve takeaways from the articles I read (like this one).

Of course sometimes you just want entertainment, in which case this isn't worthwile.

How to Work Hard 5 years ago

The attitude of feeling guilty when you're not working can be useful to motivate yourself, but I think it s also something to be careful of.

Indeed, sometimes this time of pressure can grow so pervasive it becomes impossible for you to relax, and humans aren't endlessly working robots: we need to also have time where we calm down, which PG explained.

However, he didn't warn of developing too much of a work ethic to the point relaxation itself is something you don't enjoy.

- "The Theoretical Minimum: What You Need to Know to Start Doing Physics" by Leonard Susskind. I'm reading the entire series and it's very clear and interesting, definitely recommend if you want to get into physics.

- "Speech and Language Processing (3rd ed. draft)" - Really good intro to natural language processing.

- Kar by Orhan Pamuk - not a technical book just an intriguing fiction book in turkish I want to read to work on my Turkish;.