ilia-a

5 ms·

1.
by ilia-a 15d ago
Unfortunately at the moment the model is very quickly burning through plans, single session with 3-4 subagents, none using Max or Xhigh, mostly medium + some High can burn through 5h limit within 20-40 minutes of usage.
2.
by ilia-a 3mo ago
DuckDB is really neat, recently added PDO interface for it for PHP https://github.com/iliaal/pdo_duckdb Still a bit raw, but getting there
3.
by ilia-a 3mo ago
First of all there is a way to made agents spot truncation by being aware of RTK compression and having bypass option (I use RTK_DISABLE=1) as a way of restoring original full text. Works fine, yeah it only compresses command output so only
4.
by ilia-a 3mo ago
PHP
5.
by ilia-a 3mo ago
I did fill the form our a while back (it was around for a few months now) without any response. I guess must be really big OSS project for maintainer to qualify.
6.
by ilia-a 4mo ago
Seems odd limit, especially since it highly dependant on Token provider used, with Opus this is not much and could easily be burnt in a week or less, but with something like deepseek the 1500 can literarily be an annual budget. That being s
7.
by ilia-a 4mo ago
I mean they are comparing themselves to Haiku of all things, geez that's not a good start...
8.
by ilia-a 4mo ago
Because of the memory costs lately, I doubt this will be much cheaper. Also this is quite a bit slower than even 4070 let alone *90 Nvidia variants albeit with much lower memory.
9.
by ilia-a 4mo ago
Looks like just rebranded DGX in laptop form, the biggest miss is the weak memory speed, 1/2 of the M5 laptop memory speed, and 1/3 of the M3 ultra that is now years old...
10.
by ilia-a 4mo ago
Yeah, it is a massive jump in price, hardly a "Flash" model anymore... I wonder if they'll release flash lite or something with a bit more affordable price point.
11.
FastChart: Native PHP extension for server-side charts, barcodes, and QR codes(github.com)
1 points by ilia-a 4mo ago|1 comments
12.
by ilia-a 4mo ago
Native PHP extension: 19 chart types behind a fluent OO API, plus Code 128 barcodes and QR codes sharing the same gd substrate. Composable: charts can draw into a caller-owned GdImage canvas, so multiple charts composite onto one image in a
13.
by ilia-a 4mo ago
IMHO Mythos was more of a marketing ploy. When it comes to security and AI, all top tier publicly accessible models (GPT 5.5, Opus 4.7) and even near-top like Deepseek 4 PRO can do a very good job given detailed harness on how to spot issue
14.
by ilia-a 4mo ago
Looks like lots of grift going on. Forgetting other categories for a moment, the fact that corp operations are nearly 2x the money going to linux kernel says it all...
15.
by ilia-a 4mo ago
I wonder how much shared hosting is there really left, I imagine much of it move to VPS or cheap cloud boxes.
16.
by ilia-a 4mo ago
Original online communities (forums/chat rooms) got largely killed by social media, now social media is getting killed by AI slop.
17.
by ilia-a 4mo ago
Another nail in the web anonymity sounds like
18.
by ilia-a 4mo ago
Interesting that the 5h limits are raised, but if I understand announcement correctly, the weekly limit is not. So all this means is that you can burn through your weekly limit faster and be locked out entirely, or having to buy tokens
19.
Whetstone: AI agents don't lack capability, they lack process(ilia.ws)
2 points by ilia-a 5mo ago|0 comments
20.
by ilia-a 5mo ago
Even without writing code LLMs are a huge help, analyzing code, doing code reviews, documenting code, etc... Even without writing a line of "code" LLM hugely speed up development and take away the annoying/boring work.
21.
by ilia-a 5mo ago
Compared to what was before and after it was pretty great.
22.
by ilia-a 5mo ago
Windows 10, especially when they added WSL2 was genuinely great and an awesome platform for development, with essentially native/fast Linux. Since Win11, it has all been downhill (imo), speed dropped massively, even opening task manage
23.
by ilia-a 5mo ago
Seems kinda obvious, given that most large recruiting firms/hr use algos to analyze resumes and AI written version likely do a better job at hitting keywords/structure algos/llms pick up on...
24.
by ilia-a 5mo ago
That doesn't seem right and seems to miss GLM 5.1 and Kimi 2.6. Not to mention there is the whole argument of cost/value that Chinese OSS models have vs GPT/Claude.
25.
by ilia-a 5mo ago
I use codex plugin to often validate claude results, surprisingly many findings, so for things like code-review becomes very helpful. Considering trying using Deepseek 4 Pro, seems like could be a good/cheap alternative to codex
26.
by ilia-a 5mo ago
Not surprising, hit my 5h limit on Claude Code Max Plan, had some credits so switched to extended (api). 40 minutes later $30 credits gone... so yeah, I can see how this can happen.
27.
by ilia-a 5mo ago
Yeesh, first kernel root exploit, now this... Low key wonder if people using LLMs to scan these old code bases for corner case issues and fining treasure troves of exploits.
28.
by ilia-a 5mo ago
Silly move since combo of skills/agents can achieve same results on most recent models anyway