starchild3001

5 ms·

31.
by starchild3001 1y ago
Here's some heavy research for you -- Model 3 is competing with the likes of BMW, Audi etc. That's not considered the "affordable" tier. It's called luxury. Here's a comparison: https://www.truecar.c
32.
by starchild3001 1y ago
Feels like Musk should step down from the CEO role. The company hasn’t really delivered on its big promises: no real self-driving, Cybertruck turned into a flop, the affordable Tesla never materialized. Model S was revolutionary, but Model
33.
by starchild3001 1y ago
I get the sentiment, but I think declaring scaling "dead" is premature and misses the point. First, let's be honest about GPT-5. The article cherry-picks the failures. For 95% of my workflow -- generating complex standalone c
34.
by starchild3001 1y ago
My hunch: we’ll see three things happen in parallel - AI backend providers vertically integrating into energy production (like xAI’s gas plants, or Meta’s local generation experiments), - renewed interest in genuinely efficient computing pa
35.
by starchild3001 1y ago
An MIT study concluded that only about 5% of news articles show genuine subject expertise and report facts responsibly. The other 95% are primarily driven by clickbait—optimizing for attention and engagement at the cost of accuracy—leaving
36.
MIT Viral Study Debunked [video](youtube.com)
3 points by starchild3001 1y ago|1 comments
37.
by starchild3001 1y ago
This feels like a real inflection point for image editing models. What stood out to me isn’t just the raw generative quality, but the consistency across edits and the ability to blend multiple references without falling apart. That’s someth
38.
by starchild3001 1y ago
Hypothesis: If you had ~1M dollar to burn, I think we should try setting up an AI agent to explore and try to invent new mathematics. It turns out agents can get an IMO gold with Gemini 2.5 Pro production model only. Therefore I suspect a s
39.
by starchild3001 1y ago
I love this direction of research. Reducing costs of reasoning is a huge ongoing challenge in LLMs. We're spending so much energy and compute resources today on reasoning that today's consumption rates were unexpected (to me) a
40.
by starchild3001 1y ago
What I like about this post is that it highlights something a lot of devs gloss over: the coding part of game development was never really the bottleneck. A solo developer can crank out mechanics pretty quickly, with or without AI. The real
41.
by starchild3001 1y ago
I think this essay lands on a useful framing, even if you don’t buy its every prescription. If we zoom out, history shows two things happening in parallel: (1) brute-force scaling driving surprising leaps, and (2) system-level engineering f
42.
by starchild3001 1y ago
Re: AI. I believe this will still be a human operation, as far as I can see. Awesome demo! It's a little bit of middle age crisis :), but superbly done! Thank you.
43.
by starchild3001 1y ago
I mostly follow Amiga and C64 (a little bit). I don't follow the platforms you're talking about.
44.
by starchild3001 1y ago
Re: Online content. I'm well aware of what's available out there as online content (it's no farther than a Google or youtube search). Do you think what's out there as online content is what's truly possible if we ha
45.
by starchild3001 1y ago
Amiga forever! Curation first. Two follow ups: 1) An amiga museum with all the games, artwork, coding technology, music technology etc. Perhaps an AI can be tasked to produce all of this soon. Youtube videos might be an engaging delivery me
46.
by starchild3001 1y ago
> You seem to be speaking about your own link, not the linked article on Hacker News. Correct. The hacker-news linked article talks about drugs that made mice live longer, my article talks about drugs that made humans live longer wit
47.
by starchild3001 1y ago
This is a great discussion on longevity, though the main article focuses heavily on mouse studies. To add to that, I've been thinking about a framework that prioritizes the existing human evidence. My take, which I wrote about in t
48.
by starchild3001 1y ago
Correct. People claim these models "saturate" yet what saturates faster is our ability to grasp what these models are capable of. I, for one, cannot evaluate the strength of an IMO gold vs IMO bronze models. Soon coding capabiliti
49.
by starchild3001 1y ago
If you focus on the year over year jump, not on absolute numbers, you realize that the improvement in public test isn't very different from the improvement in private test.
50.
by starchild3001 1y ago
A few data points that highlight the scale of progress in a year: 1. LM Sys (Human Preference Benchmark): GPT-5 High currently scores 1463, compared to GPT-4 Turbo (04/03/2024) at 1323 -- a 140 ELO point gap. That translates into
51.
by starchild3001 1y ago
I’m baffled by claims that AI has “hit a wall.” By every quantitative measure, today’s models are making dramatic leaps compared to those from just a year ago. It’s easy to forget that reasoning models didn’t even exist a year back! IMO Gol
52.
by starchild3001 1y ago
> This assessment fits with my anecdotal evidence. LLMs just cannot reason in any basic way. LLM reasoning is brittle and not like human cognition, but it is far from zero. It has demonstrably improved to a point where it can solve compl
53.
by starchild3001 1y ago
At this rate, the debt ceiling’s going to need its own space program! Thank you, Mr. T, and to all those before you who let avoidable crises happen, handed out tax cuts to your richest friends, and slapped it all on the national credit card
54.
by starchild3001 1y ago
It reminds me of the early-2010s SaaS wave. Back then, you could argue Salesforce or ServiceNow had huge TAMs and defensible positions, but the best entry points were when their multiples compressed and execution caught up to expectations,
55.
by starchild3001 1y ago
> Are all these post/mid-training tweaks important with abundant, verified, synthetic domain data? No. Many are aimed at cleaning/aligning noisy, mixed-domain data. With abundant, high-quality domain data, you can skip most of
56.
by starchild3001 1y ago
The reality is that cloud and AI model serving has become a race to the bottom. Top models are increasingly similar in capabilities, so price, not features, has become the main differentiator. OpenAI’s key move was to put its thinking capab
57.
by starchild3001 1y ago
Yes, I occassionally use LLM for edits and re-writes. Opinions are mine. I thought most people do these days? (Re: Other post you linked to. it is entirely my own thoughts.)
58.
by starchild3001 1y ago
Really appreciate the depth of this paper; it's a welcome change from the usual model announcement blog posts. The Zhipu/Tsinghua team laid out not just the 'what' but the 'how,' which is where the most interes
59.
by starchild3001 1y ago
What stood out to me is how much of gpt-oss’s “newness” isn’t about radical architectural departures, but about a careful layering of well-understood optimizations—RoPE, SwiGLU, GQA, MoE—with some slightly unusual choices (tiny sliding-wind
60.
by starchild3001 1y ago
Glad to see "dead internet theory" not holding up! PS: As someone who worked on internet search, I can assure you that at least half of most popular web pages change in about 6 months time. And the change is in no way something th

More ›