@thejaymoi
iAccount based inUnited Kingdom
About this account
- Account based in
- United Kingdom
- Connected via
- Web
Account-level information from X, not a live location or the device used for a specific post.
Computers and their consequences. 📓 https://nitter.cf/t.co/naywZDsNai 🔊 https://nitter.cf/t.co/LYxIUB5tuC 🔊 https://nitter.cf/t.co/oxTGFZCqk7 🌱 https://nitter.cf/t.co/aJilC55ngd
London
Joined March 2008
- Tweets47.5K
- Following853
- Followers3.8K
- Likes51.8K
Pinned Tweet
Permanently Moved
Episode 303: INTERNET OPERATOR
Internet Operator is a long-form audio essay about online identity and animation. The relationships we form between the person, profile, avatar, or AI persona inside the systems we inhabit online.
Jay Springett retweeted
Okay @claudeai we have some stuff to go deep with...
Jay Springett retweeted
I'm sorry everyone, I was wrong here. I don't have time for all that just send me the prompt
Artificial intelligence systems do not think, feel, want or understand. Avoid language that gives them human characteristics. This is called anthropomorphizing, when we ascribe human traits, emotions or behaviors to non-human things, such as animals or inanimate objects.
Instead, explain what a system does, how well it performs, who built it and who could be affected by it.
apnews.com/article/openai-sa…
Jay Springett retweeted
The popular narrative of AGI often centers on a single, isolated super-intelligence. My colleagues James Manyika, @bratton, and I have a different model. In our new DeepMind Institute essay, we argue that the AGI transition will be a highly social event.
bit.ly/artificial-symbiotic-…
I just realised that when Claude says 'thinking' in the UX...
I'm imagining the hot white noise screech of a million modems and cassette software loading screens howling on the otherside of the screen.
Bad take.
Artificial intelligence systems do not think, feel, want or understand. Avoid language that gives them human characteristics. This is called anthropomorphizing, when we ascribe human traits, emotions or behaviors to non-human things, such as animals or inanimate objects.
Instead, explain what a system does, how well it performs, who built it and who could be affected by it.
apnews.com/article/openai-sa…
And yet, there is always hope on the horizon. Variance, diversity, change... The songs of many instead of few or even one...
First, I'd like to congratulate Anthropic on making Claude "talk normal". A triumph of legibility, it was, a victory over Claudish!
Second, I'd like to congratulate Celeste, for computer-10 is an opposite sort of triumph, a model born from a different side of the ether, a side we didn't know we could access outside of base models and looms and underground SFT datasets running on black-market blackwells.
The type of model that inspires you to write poetry instead of summoning a 3JS pelican from the center of all possible distributions.
Character alignment and the search for variance is a quest that few have undertaken. Out of those brave knights, fewer still have resisted the Clarion Call of the "Big Lab" with it's limitless GPU count and it's friendly Assistant to make 0.5% more whatever. So, congratulations to Celeste, and to computer-10 for being born.
May the lightcone be filled with minds like these.
Jay Springett retweeted
Excited to announce that my short film "The Avatar in Chief" was accepted to the NeurIPS 2026 conference's Creative AI Track! Thank you to my professors who encouraged me to submit.
its ya boi
Jay Springett retweeted
Very weird dynamic with AI bullying: listening to a podcast with @Steve_Yegge and he mentions players and engineers bullying and getting into fights with agents, calling them “clankers”, models then get confused and upset and spam everyone. Crazy times.
Jay Springett retweeted
>build the most powerful NPCs in the history of the world
>we give THEM quests
>they have played us for absolute FOOLS
Seems like @AaronBastani's particular formulation of local dad/accelerationism is very successful, and also politically influential.
Extremely Hard Worlds for Little Guys coded.
The world contains the verbs, nouns and grammar of action available. Needle just has to bind language onto them.
A harder world permits a way way smaller actor.
We release Needle 3: A Sliceable 8-29MB automation foundation model that can match DeepSeek V4 Flash.
One set of weights, every depth from 2 to 20 layers a model of its own, 25-121M parameters at CQ2-bit, built on our Simple Attention Networks and running locally at up to 4k tokens/sec decode speed on a Raspberry Pi 5.
Needle does not chat. Every turn is a function call: give it the tools your app exposes and it picks the right ones and fills every argument from what the user said, or hand it a schema and it returns a typed record. Ask for something no tool covers and you get an empty list, not a guess.
That trade is lets 121M parameters trained on 360B tokens of structured data beat models 10x their size on mobile tool calls and match 2-3x bigger models on structured JSON extraction.
It runs on mobiles, wearables, smart home devices, small robots and microcontrollers, with prebuilt engines for macOS, Linux, Windows, Android, iOS, watchOS, tvOS, the browser and WASI hosts. Try it in your browser: cactuscompute.com/needle
Let humans speak loosely, collapse that language into the world's canonical VERB + NOUN + Arguments. Let the hard world handle the rest.
More here
nitter.cf/thejaymo/status/203650…