「Ian」⛩

971 posts

「Ian」⛩ banner
「Ian」⛩

「Ian」⛩

@iantimmis

Machine Learning Engineer | Stochastic Parrot | e/acc

Katılım Ağustos 2011
769 Takip Edilen145 Takipçiler
「Ian」⛩
「Ian」⛩@iantimmis·
@leerob @cursor_ai @leerob If it helps - one failure I am getting very often is that the chat is property connected over ssh, but can't create terminal window, it hangs in loading state
English
0
0
0
6
「Ian」⛩
「Ian」⛩@iantimmis·
@cursor_ai @leerob Loving the new Cursor Glass update preview, it's everything I've been wanting. One big issue I am having - workspaces that are on different computers through SSH constantly disconnect and there is no button or setting to trigger reconnect manually.
English
1
0
0
42
「Ian」⛩
「Ian」⛩@iantimmis·
@leerob @cursor_ai Thanks for the response! This is a killer feature for me, being able to control agents on all the computers I interact with
English
0
0
0
9
Dev Ed
Dev Ed@developedbyed·
Making an app in electron to generate your own game assets,tiles,animations, inpainting and more. No subscriptions, all run locally
English
180
330
4.9K
286.2K
「Ian」⛩ retweetledi
Dylan Patel
Dylan Patel@dylan522p·
Dylan Patel tweet media
GamersNexus@GamersNexus

@PatrickMoorhead If you used tokens and AI, you did not, in fact, "personally build" it. Millions of people whose work was stolen built it. You benefited from it.

ZXX
11
15
454
33.4K
Bryan Johnson
Bryan Johnson@bryan_johnson·
2-3 cups of coffee per day associated with an 18% lower dementia risk and better cognition in a 40-year study of 130,000 participants. Even with these benefits, I can't do it; caffeine creates a physiological roller coaster and enslaves me.
Bryan Johnson tweet media
English
1.1K
276
10.3K
1.6M
「Ian」⛩ retweetledi
Alex Prompter
Alex Prompter@alex_prompter·
🚨 RAG is broken and nobody's talking about it. Stanford just exposed the fatal flaw killing every "AI that reads your docs" product. It's called "Semantic Collapse", and it happens the moment your knowledge base hits critical mass. Here's the brutal math (and why your RAG system is already dying):
Alex Prompter tweet media
English
366
679
4.9K
1M
Andrej Karpathy
Andrej Karpathy@karpathy·
I am unreasonably excited about self-driving. It will be the first technology in many decades to visibly terraform outdoor physical spaces and way of life. Less parked cars. Less parking lots. Much greater safety for people in and out of cars. Less noise pollution. More space reclaimed for humans. Human brain cycles and attention capital freed up from “lane following” to other pursuits. Cheaper, faster, programmable delivery of physical items and goods. It won’t happen overnight but there will be the era before and the era after.
English
793
2K
21.8K
1.5M
「Ian」⛩
「Ian」⛩@iantimmis·
@daniel_mac8 Pretty sure cursors new model is a fine tune of a Chinese model
English
0
0
0
17
Dan McAteer
Dan McAteer@daniel_mac8·
Honest question: why do more companies not use Chinese open weights LLMs in production? If they are at parity, yet cheaper, why wouldn’t more use them? Doesn’t make sense.
English
330
23
740
175.2K
🇺🇦 Dzmitry Bahdanau
🇺🇦 Dzmitry Bahdanau@DBahdanau·
can someone explain to me int4 training by @Kimi_Moonshot ? does it mean weights are stored in int4 and dequantized on the fly for futher fp8/bf16 computation? or does this mean actual calculations are in int4, including accumulators?
English
16
5
162
140K
Rakshit (chessiro.com)
Rakshit (chessiro.com)@Ra1kshit·
Super interested to put together a group of people that understand chess and tech well. Comment below if you are one, or i know someone in this intersection.
English
532
32
1.3K
115.6K
Namanyay
Namanyay@NamanyayG·
i'm disrupting the vibe coding space & giving out the best Claude Code models for FREE yes, FREE as in free beer. the pricing in this space has gotten prohibitively expensive creating a huge barrier of entry. with us, just 1 line of change and you get free Claude Code right in your terminal oh and it's not burning any vc money lol, it's supported by some great companies in SF 🥰 how it works: 1. you change Claude Code env variables to use our inference endpoint 2. to support the free usage, we'll add ads inside the AI responses. Only contextual and relevant ads, no spam. 3. it's not ready for public release, for now, it'll be first come first serve. comment and retweet to join the waitlist and see how it works in the video:
English
766
475
2K
273.4K
「Ian」⛩
「Ian」⛩@iantimmis·
@gdb Can ChatGPT interact with the app or is it only embedding it for the user to interact with
English
0
0
0
122
Greg Brockman
Greg Brockman@gdb·
ChatGPT Apps are very powerful and can include full-fledged applications:
Guillermo Rauch@rauchg

Can ChatGPT run Doom? Yes. ChatGPT Apps are very powerful. I cloned our Next.js ChatGPT template, registered a 𝚙𝚕𝚊𝚢_𝚍𝚘𝚘𝚖 MCP tool and deployed to @vercel. Once the tool is called, ChatGPT embeds the full @nextjs application. Server and client rendering just works, and it's 100% interactive. h/t @andrewqu @allenzhou101 for the starter kit: vercel.com/templates/ai/c…

English
60
75
1K
172.1K
「Ian」⛩ retweetledi
Theo - t3.gg
Theo - t3.gg@theo·
The best decision OpenAI made with Sora was making it iOS only. Android users lack taste and would ruin the whole thing
English
437
88
4K
350.7K
Kelly Greer
Kelly Greer@kellyjgreer·
who wants a printed copy?
Kelly Greer tweet mediaKelly Greer tweet media
English
832
431
9.7K
765.4K
difficultyang
difficultyang@difficultyang·
TIL, RIP Triton, killed by inability to have good Blackwell performance
English
18
10
375
155.3K
「Ian」⛩ retweetledi
Francesco Bertolotti
Francesco Bertolotti@f14bertolotti·
Choosing the right data mix is as crucial as hard. Luckily, his paper proposes parametric scaling laws, fitted on small settings, then solves an optimization problem to find near-optimal mixtures. Experiments show the problem is scale-invariant. 🔗arxiv.org/pdf/2508.11953
Francesco Bertolotti tweet mediaFrancesco Bertolotti tweet mediaFrancesco Bertolotti tweet mediaFrancesco Bertolotti tweet media
English
0
3
28
1.5K
「Ian」⛩ retweetledi
Ege
Ege@egeberkina·
The first iPhone without a camera! This iPhone 17 Pro spec ad goes hard. Nano Banana really solves a lot of video production challenges in different ways. Credit: Simon Meyer & Ciro Kavouras
English
43
148
1.3K
329K
「Ian」⛩
「Ian」⛩@iantimmis·
@DanielRensch I'm 3100 puzzle rating now and feel I could reach 4-5 thousand easily. Can we fix this
English
0
0
0
13
「Ian」⛩
「Ian」⛩@iantimmis·
What the heck happened to puzzles on @chesscom? I had to work hard, solve 40k+ puzzles to get >2600 With the new difficulty changes I am now >3000 with little effort I feel like my hard earned rating here is completely meaningless now, and, im learning less and having less fun
English
1
0
1
41