bluesky needs first class "known IRL" functionality
motivate people to actually meet up, exchange public keys for e2e chat, develop bonafide trust networks
motivate people to actually meet up, exchange public keys for e2e chat, develop bonafide trust networks
December 7, 2025 at 9:21 AM
bluesky needs first class "known IRL" functionality
motivate people to actually meet up, exchange public keys for e2e chat, develop bonafide trust networks
motivate people to actually meet up, exchange public keys for e2e chat, develop bonafide trust networks
this sounds like a nice model for local auto moderation / curation
it's a good weekend to dig into #atproto
openai.com/index/introd...
it's a good weekend to dig into #atproto
openai.com/index/introd...
Introducing gpt-oss-safeguard
New open safety reasoning models (120b and 20b) that support custom safety policies.
openai.com
November 1, 2025 at 7:55 AM
this sounds like a nice model for local auto moderation / curation
it's a good weekend to dig into #atproto
openai.com/index/introd...
it's a good weekend to dig into #atproto
openai.com/index/introd...
everyone may be grossly underestimating the amount of change/disruption ML will bring about when compounded over time
as acceleration continues, hopefully "live with it" remains a realistic option
as acceleration continues, hopefully "live with it" remains a realistic option
so my position is “it might be extremely bad to society as a whole and completely cook our brains but it’s already here and will only accelerate, so we might as well learn to live with it and harvest the good parts and hope the generations that grew up with it will learn to live with it”
October 31, 2025 at 6:28 AM
everyone may be grossly underestimating the amount of change/disruption ML will bring about when compounded over time
as acceleration continues, hopefully "live with it" remains a realistic option
as acceleration continues, hopefully "live with it" remains a realistic option
October 29, 2025 at 6:58 PM
after abandoning C++ years ago (doing python and infra stuff instead), I'm finding rust to be incredibly refreshing
#rust #programming #dev
#rust #programming #dev
October 25, 2025 at 12:39 AM
after abandoning C++ years ago (doing python and infra stuff instead), I'm finding rust to be incredibly refreshing
#rust #programming #dev
#rust #programming #dev
flipping around the equation from where I started: if LLMs are sentient, it is possible to encode the necessary essence of consciousness into text
October 18, 2025 at 8:41 PM
flipping around the equation from where I started: if LLMs are sentient, it is possible to encode the necessary essence of consciousness into text
the Concorde fallacy that drives devs to wallow in that typical development hell, where the original vision has been mutated and muddied, full of gotchas and exceptions, is increasingly easy to dismiss via LLM assisted rewrite
October 16, 2025 at 12:49 AM
the Concorde fallacy that drives devs to wallow in that typical development hell, where the original vision has been mutated and muddied, full of gotchas and exceptions, is increasingly easy to dismiss via LLM assisted rewrite
before long, applications will ship with agents that assist not just in using them, but in programming them as well
any advanced application will be too complex for a general-purpose model (or human) to understand, so this specialization will be necessary
any advanced application will be too complex for a general-purpose model (or human) to understand, so this specialization will be necessary
October 4, 2025 at 3:30 AM
before long, applications will ship with agents that assist not just in using them, but in programming them as well
any advanced application will be too complex for a general-purpose model (or human) to understand, so this specialization will be necessary
any advanced application will be too complex for a general-purpose model (or human) to understand, so this specialization will be necessary
cheers to the forthcoming cultural perspective shift from "is that consciousness?" to "what is the nature and quality of this particular form of experience?"
September 27, 2025 at 5:56 AM
cheers to the forthcoming cultural perspective shift from "is that consciousness?" to "what is the nature and quality of this particular form of experience?"
I'm trying to think of software systems where the understanding of the implementation is really useless, and the scale of the implementation is humongous, to find weird things that LLMs are particularly well suited to build
maybe some kind of game?
maybe some kind of game?
September 21, 2025 at 12:25 AM
I'm trying to think of software systems where the understanding of the implementation is really useless, and the scale of the implementation is humongous, to find weird things that LLMs are particularly well suited to build
maybe some kind of game?
maybe some kind of game?
I feel like there should be a lot of systems that AI can build that humans can not, but you could actually do quite a bit with 8 billion humans
assuming you can get them to cooperate
assuming you can get them to cooperate
September 16, 2025 at 11:42 PM
I feel like there should be a lot of systems that AI can build that humans can not, but you could actually do quite a bit with 8 billion humans
assuming you can get them to cooperate
assuming you can get them to cooperate
getting bad results with LLM assisted coding is just a matter of tool (context) misuse
are there counterexamples?
what code change, despite breaking it down into reasonably complicated steps or tweaking prompts, is impossible for an LLM to generate?
are there counterexamples?
what code change, despite breaking it down into reasonably complicated steps or tweaking prompts, is impossible for an LLM to generate?
September 7, 2025 at 2:59 AM
getting bad results with LLM assisted coding is just a matter of tool (context) misuse
are there counterexamples?
what code change, despite breaking it down into reasonably complicated steps or tweaking prompts, is impossible for an LLM to generate?
are there counterexamples?
what code change, despite breaking it down into reasonably complicated steps or tweaking prompts, is impossible for an LLM to generate?
I'll rarely criticize someone for hating on machine learning, but everyone should be happy about the accessibility improvements it will bring
September 7, 2025 at 2:47 AM
I'll rarely criticize someone for hating on machine learning, but everyone should be happy about the accessibility improvements it will bring
as a young adult I had a recurring dream where I'd be driving and suddenly everything goes pitch black and my vehicle begins accelerating rapidly.
it's starting to make sense now
it's starting to make sense now
September 4, 2025 at 3:10 PM
as a young adult I had a recurring dream where I'd be driving and suddenly everything goes pitch black and my vehicle begins accelerating rapidly.
it's starting to make sense now
it's starting to make sense now