daquisu
Joined 4 years ago
Karma
49
Submissions
30
Member since
Nov 30
Recent submissions
Comment · 1 month ago
Comment · 1 month ago
There are some users reporting it improved a lot in the last few weeks. The max sub usage for Grok is around $12,000 of API pricing now, so a 40x multiplier for the $300 plan.
It is the same multiplier for Sol with subscription. For Astra though the multiplier is ≈20x, so half of Sol usage.
For Claude it seems to be ≈40x too for Opus, but less for Fable (similar to Astra in GPT).
All on the most expensive plan. Previously, Grok usage escalated linearly from the $100 plan to $300 plan. That would be a really good $100 plan if it is still true.
Some sources:
1. https://x.com/kunchenguid/status/2098256018836963382
2. https://x.com/stevenzhang/status/2092110386569089311
3. https://github.com/openai/codex/issues/43731
Comment · 1 month ago
That is even worse. Back to 5th grade I guess
Comment · 2 months ago
I read the full post. It is more about cheap models with good enough performance rather than small models. "Local" was not mentioned at all
Comment · 2 months ago
Related:
- Ozempic For Sleep https://www.isaak.net/sleepless/ (November 2024)
It is the same author from https://news.ycombinator.com/item?id=41742432
He recently founded https://capable.com, also related
Comment · 4 months ago
It was a recent edit though. Yesterday snapshot: https://web.archive.org/web/20260613072958/https://huggingfa...
Comment · 4 months ago
Later in the same blog post, the author says:
> We can also consider the IMO 2025 problems individually. In the Epoch AI newsletter, Greg Burnham combines a subjective analysis with Evan Chen’s MOHS ratings to argue that the first five problems at IMO 2025 were unusually easy and the sixth was unusually hard, so it’s not surprising that the first five problems were exactly the ones solved by these AIs. Though I’m not sure the MOHS scale is rigorous enough to make sense as the x-axis of a bar chart it’s easy to corroborate the high-level story with the official IMO statistics. Based on average scores, this year’s Problem 6 was the fourth hardest and its Problem 3 was by far the easiest of all Problem 3s and 6s since 2000.
In the linked MaxProof paper, in the section "6.3.1. Per-Problem Analysis" it shows the same behavior: 7/7 in the first 5 problems, 0/7 in the last problem.
Comment · 4 months ago
"I thought it was interesting and a bit underappreciated that the fraction of gold medalists at the 2025 IMO (72/630 = 11.4%) is the highest it’s been since 1981.
Crudely, IMO gold medals are awarded to the highest-scoring 1/12 of contestants.1 However, because scores are integers up to 42 and there’s no provision for tiebreaking, it’s possible for a lot of contestants to be tied around the threshold. In that case, either all of them get a gold medal or none do, and the fraction of gold medalists might deviate substantially from 1/12. That’s what happened this year: 46 contestants all won a gold medal by scoring exactly 35 points.
In fact, bizarrely, 35 is the mode of the scores this year; the last time the modal score was a gold medal score was in 1994. And, of course, 35 is the same score claimed by AI systems from Google, OpenAI, and others."
Comment · 4 months ago
I just tried that since I read your comment and it is really helping me. Thanks!
Comment · 10 months ago
Now it is even easier. Cloudflare has a beta product called AI Search that implements most of these 160 lines of code
Comment · 10 months ago
12.5 million a year for a hundred people seems reasonable? 125k per person per year. GP still said "a few hundred" - two hundred would drop that value to 62.5k per person
Comment · 1 year ago
Firefox on mobile works with ublock. It can also play videos even with the screen locked, although you do have to unpause it after locking the screen.
Comment · 1 year ago
> The "You are an expert software engineer" really helps?
Anecdata, but it weirdly helped me. Seemed BS for me until I tried.
Maybe because good code is contextual? Sample codes to explain concepts may be simpler than a production ready code. The model may have the capability to do both but can't properly disguished the correct thing to do.
I don't know.
Comment · 1 year ago
That is a common narrative but Google had LaMDA as an LLM with over 100B parameters before the ChatGPT release. There was even a Xoogler that claimed it was alive.
From my POV Google could have released a good B2C LLM before OpenAI, but it would compete with their own Ads business.
I can't help much more than that, I did that research in the last few days, but I never used Grok myself.
I pay for GPT, Claude and Gemini. Last week I consumed all my quota on two of them, so I wondered which next subscription I would pay for if needed.