senko

18 ms·

31.
by senko 2mo ago
Of course you can beat it (in this context). You can cheat (by shifting the work around). Started to write up a comment, got a bit wordy, turned into a blog post: https://senkorasic.com/articles/ai-amdahls-law
32.
by senko 2mo ago
> What would this do to the insurance or pension firms who hold this debt? If the debt is serviced, nothing.
33.
by senko 2mo ago
So, what happens when it stops in this particular instance ? The article is about hyperscalers which are massively profitable irrespectively of AI. If the lending stops and this leads to paused or cancelled infra projects, but these compan
34.
by senko 2mo ago
I've been researching the alternatives as I've been really.annoyed by the Meetup.com enshittification. Any other platforms besides Luma you hear about? I know some lean on WhatsApp, some tried to use FB or even LinkedIn events but
35.
by senko 2mo ago
Now, see, when AI does something like this, we call that "hallucination". Thomas explained what he means in a sibling comment an hour before yours.
36.
by senko 2mo ago
Wow, what an epic misread :)
37.
by senko 2mo ago
Azul[0] is a great board game inspired by this style. Easy to start, can get quite complex, heartily recommended! [0] https://boardgamegeek.com/boardgame/230802/azul
38.
by senko 2mo ago
I still don't understand why tmux authors wasted their time on tmux, tho, when we already had perfectly servicable GNU screen. Why reinvent the wheel? /s
39.
by senko 2mo ago
Article author here, thanks for sharing! I mainly wrote this as a reflection on many accounts of OSS maintainers burning out. Much has changed since 2016, but I believe the point still stands. The arrival of AI based coding (and Pull Reques
40.
by senko 2mo ago
They might be rare, but they do exist: https://en.wikipedia.org/wiki/Igalia
41.
by senko 2mo ago
Not an expert, but looks like they did a lot more work on the RL part (9 expert models, full sandbox access for agentic tasks, etc)?
42.
by senko 2mo ago
Old but relevant: if you read the recently-released Kimi K3 paper[0], you'll see that it's heavily based on Kimi Linear discussed here, scaling it up and adding a bunch more things (like native vision and RL improvements). [0] ht
43.
by senko 3mo ago
You mean like GitHub?
44.
by senko 3mo ago
It's hilarious. The Economist often has satire columns, mostly related to office or biz absurdities.
45.
by senko 3mo ago
> Being left handed is fine As a lefty, I would rather never write another word then twist my hand the way he's doing (above the current line) to avoid the smudges.
46.
by senko 3mo ago
> The Ketgpt takes the kettle into the agentic era. At the front door and fancy a cup of tea? Just woken up and desperate for your first coffee? Get out your phone, fire up the Ketgpt app, enter your login details and tell your kettle t
47.
by senko 3mo ago
FWIW my observation was about the models I tested, but I can see how it could be taken as a general statement.
48.
by senko 3mo ago
For the web app task I mentioned: * Kimi K3: 9532k input (9172k cached), 114k output - cost $5.5 * Qwen 3.8 Max: 18020k input (17836k cached), 114k output - cost $6.3 * Fable: ~14m input (all cached??), 196k output - cost $30 Correction on
49.
by senko 3mo ago
Having tested K3, Qwen 3.8 max preview, Fable and Sol for the past few days, I partially agree. Don’t trust the benchmarks, and the Chinese models really are slow and token-inefficient. However they do seem very close to SOTA : I’d say roug
50.
by senko 3mo ago
Here it is: https://senko.net/vibecode-bench/2026/rts-gpt-5.6-sol.html Clearly much better than the Terra version. I'd say its on par with Fable, and the observed differencies are more due to random luck and
51.
by senko 3mo ago
Here's the prompt I used: > Create a simple but functional real time strategy (RTS) game similar to old WarCraft, StarCraft or Command & Conquer games. The player should be able to build buildings, create units, gather resource
52.
by senko 3mo ago
I'd say Fable 5: https://senko.net/vibecode-bench/2026/rts-fable-5.html It even has enemies! (I'm not too mad about it not following my instructions because it can be fun to play :) And I generated that
53.
by senko 3mo ago
Well, it is a silly test, not a scientific benchmark. However, I would say it is a measure (not the measure). If you look at the entries, there's a lot of variation - definitely not something they memorized outright. And the test its
54.
by senko 3mo ago
You guys had filesystems??
55.
by senko 3mo ago
I love testing the new models by asking them to code a toy RTS game. Here's what Terra did: https://senko.net/vibecode-bench/2026/rts-gpt-5.6-terra.html (one try, in codex app, xhigh effort) Comparing this to
56.
by senko 3mo ago
'fraid so, the odds have risen too high. We might need a white night on a fiery steed. One can dream.
57.
by senko 3mo ago
> I feel like EU could start a company, That's not how market-based economies work... > feed 2bln a year into it and make a compelling almost SOTA model ...and the reason is, if you give a bunch of people €2b a year and tell
58.
You're Right(youre-absolutely-right-one.vercel.app)
2 points by senko 3mo ago|0 comments
59.
by senko 4mo ago
The post mainly talks about coding from security point of view. Fair enough. In my own (limited) testing so far, Fable is the most capable model (for coding in general), and the most expensive. It pretty much saturated my "LLMCraft&quo
60.
by senko 4mo ago
LinkedIn in particular is quite aggressively blocking any automated attempts to read or navigate through it. I post quite a lot there and wanted to have a copy of my posts on my blog[0] to preserve them. For a few months I was able to use

More ›