> This also means that languages that are not well represented online typically aren't able to be compressed well you havent referred to the idea of llms as compression anywhere prior in the post overall good article!
@swyx
-
Request for more behind-the-scenes technical content and broader coverage
By
–
the behind the scenes on claude thread you did was great, more like that anything @AmandaAskell or @benskuhn is good but also it doesnt gave to be an anthropic only newsletter. you have credibility if you can objectively discuss other stuff, including competing
-
Security and Safety Talks: Low Attendance Despite High Importance
By
–
there's a whole conf in here for security/safety as well. i feel like the supply vs demand ratio for security/safety talks are hard to gauge. i dont know how to handicap here. its like testing in trad software – everyone agrees its impt, then the talks get crickets
-
NYT’s Failed Gambit Against OpenAI: Content Uniqueness Miscalculation
By
–
the thing we talked about in the Data War is coming to pass – in breaking off negotiations and choosing the nuclear option with openai; the NYT made a hard bet on the uniqueness of their content oai simply shrugged and went with the next best paper of record. NYT management may
-
Man-in-the-Middle Attack on Network Calls and Request Headers
By
–
itd be trivial to mitm the network calls and get the right request headers tho?
-
LLMs Structured Output as Postel’s Law Implementation
By
–
LLMs that implement constrained sampling/schema respecting structured output seem the ultimate implementation of Postel's Law: "Be conservative in what you send, be liberal in what you accept." Basically the ultimate interface block between systems.
-
Strong to Weak Alignment: A New Perspective on AI Safety
By
–
tired: weak to strong alignment wired: strong to weak alignment
-
Personal AI Remembers Book Conversation and Purchases It
By
–
had conversation about book personal ai remembered that fact asked ai to buy the book without mentioning its title received book all without typing
-
TAM/SOM Discussion and Commercial Open Source Relicensing Drama
By
–
fantastic discussion of the TAM/SOM, and the tough choices of commercial open source. best solo guest to discuss the recent relicense drama!
-
Conditioning and In-Context Learning: Architecture’s Impact Analysis
By
–
what i am confused that more people dont seem to be analysing is how does the conditioning improve (or not) with increasing attention heads and layers. i dont know if anyone is formally quantifying amount of conditioning/ICL as an explicit loss function/training goal. in other