For anyone with endless ideas, this agent age is nirvana as those ideas are met with endless execution, endless exploration. I've never had has much fun working with computers as I do right now. What a time to be alive.
This morning I had to fill some customs documents. I nearly went to my local copy shop to print the documents, fill and sign everything by hand, scan them with my phone (I don't own a printer).
But I was too lazy to do it so I gave codex the required info and created a signature in apple preview.
It was done pretty quickly.
The final lecture of my course is an intro to character training! This is a topic that I've been quietly very invested in for ~18 months, as it:
* Has potential for high real world impact
* Clearly used extensively at frontier labs
* Almost no empirical literature exists
* More accessible on academic compute
This lecture covers what character training is, reviews model specs, constitutions, the differences, the motivations in real world events, some example research papers I like, and open questions in how it relates to post-training/model use generally.
Hopefully this brings more people into the field (and reach out if you have questions). It is one of the more research-y chapters in my book, but one that I felt needed the reference. There is still so little, educational content on the topic online.
0:00 Intro
6:22 Part 1: Fundamentals — character, constitutions, and model specs
19:21 Part 2: Character training in practice
23:23 Part 3: Character elicitation without gradient steps
28:03 Part 4: Open questions (and the end of the course)
32:27 The course, complete
Thanks for watching. No need to like and subscribe now that the course is done, you definitely wouldn't!
h/t to @_maiush for leading the technical work I got to do in the space, and @zafstojano for investing a lot of attention at this book chapter.
The @lokahq team post-trained our 26B Trinity Mini into an agentic harness specialized for scientific work.
Working with @PrimeIntellect and @AWS, Loka taught Trinity Mini to execute multi-step research workflows, coordinate tools, query databases, and log an auditable trail at every step.
Here is a look inside the harness. 🧵
My conversation with @dhh, co-founder of @37signals.
0:00 Cutting Rework From 50,000 Words To 25,000
2:54 Basecamp Was Built In 380 Hours
6:46 Unlimited Resources Build A Blob
10:22 Terminator Beats Avatar By A 1000x
15:24 The Printer DHH Never Wants To Upgrade
17:31 Millions In Profit From 16 Year-Old Software
20:10 The Beauty Of A Finished Product
22:54 I Would Quit Before Working In An Office
25:41 Four Uninterrupted Hours Or The Day Is Lost
31:21 Tobi Lütke Saw The AI Future First
38:51 Apple Pissed Him Off Enough To Switch
42:47 The 19 Year-Old Who Beat Apple
47:16 If Nobody Else Will Build It, I Will
52:02 The Need For Control, Independence, And Freedom
53:44 Chef's Choice, Not A Construction-Kit Burger
56:50 A New Computer Ready In 2 Minutes
1:00:33 A Cheaper Mac Will Never Win
1:02:27 The Best Ideas Sound Stupid At First
1:03:47 Out-Teach Your Competition
1:06:57 The Algorithms Broke Reciprocity
1:10:53 No Investors Is Freedom. No Customers Is Transcendence.
1:13:25 No One Gets To Tell Me What To Do
1:16:49 Your Moral Obligation To Buy A Ferrari
1:23:19 Costs Are The Only Thing You Control
1:26:49 Bezos Paid For The Promise
1:30:17 The Confidence To Say No To Everyone
1:32:59 The Day A Third Of The Company Left
1:36:49 Marc Andreessen Came To Help
Includes paid partnerships.
Friends @Shekswess and @marioxpet visiting Prague and i can’t show them the city so I compiled 72 @realHonestGuide restaurant, pub & café picks in ~10 min while walking my 🐕 in the park. Check it here: https://t.co/YAtoYELoRO
Built with @cursor_ai mobile app + @Apify from their public videos. Unofficial fan list, all credit to Janek & Honza.
Support them: https://t.co/RG9rGQCmf5