kgeist

14 ms·

181.
by kgeist 8mo ago
You can prompt an LLM to add typos, though
182.
by kgeist 8mo ago
Technically, clients could simply upload images to S3 and render them inline whenever they encounter an image URL in a message - no protocol or server changes required. And if you're using an older IRC client, you can still just click
183.
by kgeist 8mo ago
Russia has been among the top sources of spam since the early 2000s, it's not like anything changed lately. Mail-order bride scams and similar peaked in like 2005. It doesn't take a lot of people to send spam, I don't think i
184.
by kgeist 8mo ago
There was an article by Raymond Chen where he argued that giving app developers an API option to say "run me under high/low priority" rarely works because every developer views their program as the main character on the stage
185.
by kgeist 8mo ago
Windows has SetThreadPriority and SetThreadAffinityMask since at least Windows XP.
186.
by kgeist 8mo ago
Can't Windows/Linux pin background threads to specific cores on Intel too? So that your foreground app isn't slowed down by all the background activity going on? Or there's something else to it that I don't understa
187.
by kgeist 8mo ago
>Yes it's 100% prioritization Amazon Bedrock has a similar feature called "priority tier": you get faster responses at 1.75x the price. And they explicitly say in the docs "priority requests receive preferential treat
188.
by kgeist 8mo ago
Years ago I built a scripting language that transpiled to TCC and then compiled to machine code in memory. It produced human-readable C code so it was very easy to get going: when debugging the compiler I could just look at the generated C
189.
by kgeist 8mo ago
>is going to improve the human programming skills worldwide to levels never before imaginable "We found that using AI assistance led to a statistically significant decrease in mastery. On a quiz that covered concepts they’d used jus
190.
by kgeist 8mo ago
It creates a weird feedback loop: after I watch video A, it recommends a similar video B, and if I make the mistake of watching that too, it then recommends video C on the same topic. Suddenly my feed is nothing but Stranger Things shorts f
191.
by kgeist 8mo ago
>The truth is it's just as far from intelligence your music playback device is from having Elvis inside of it. A music playback device's purpose is to allow you hear Elvis' voice. A good device does it well: you hear Elvis
192.
by kgeist 8mo ago
>Unfortunately it’s as legacy as Jenkins today Why do you consider TeamCity legacy? The latest release was just 2 months ago: https://www.jetbrains.com/help/teamcity/what-s-new-in-teamci... >To make TeamCity
193.
by kgeist 8mo ago
A lot of the internet is duplicate data, low quality content, SEO spam etc. I wouldn't be surprised if 1 TB is a significant portion of the high-quality, information-dense part of the internet.
194.
by kgeist 9mo ago
ikllama.cpp is a fork of llama.cpp which specializes on CPU inference, some benchmarks from 1 year ago: https://github.com/ikawrakow/ik_llama.cpp/discussions/164
195.
by kgeist 9mo ago
>The hardest part about this project was actually just parsing. How about using sqlite for this? Then you wouldn't need to parse anything, just read/update tables. Fast indexing out of the box, too.
196.
by kgeist 9mo ago
Not denying that Russia abuses Interpol, but I have doubts about this particular narrative that he was some kind of "government critic." From what I can find, he privatized a state corporation in the 90s for pennies (lots of very
197.
by kgeist 9mo ago
Are there vector DBs with 100B vectors in production which work well? There was a paper which showed that there's 12% loss in accuracy at just 1 mln vectors. Maybe some kind of logical sharding is another option, to improve both accura
198.
by kgeist 9mo ago
>Didn't Russia quickly spin up an alternative smartcard payment system The MIR payment system started functioning in 2015, long before Visa/Mastercard pulled out of Russia >Android app store Initially there was some fragment
199.
by kgeist 9mo ago
>If no human ever used that phrase, I wonder where the ai's learned it from? Reinforced with RLHF? People like it when they're told they're right.
200.
by kgeist 9mo ago
Early LLMs used to have this often. I think's that where the "repetition penalty" parameter comes from. I suspect output quality can be improved with better sampling parameters.
201.
by kgeist 9mo ago
I wonder why Gmail and other email providers don't just run an LLM/ML pipeline to detect phishing emails. It seems that matching an email's content with the sender's domain (and possibly analyzing the content behind link
202.
by kgeist 9mo ago
I think they mean that they trained the tool-calling capabilities to skip personal information in tool call arguments (for RAG), or something like that. You need to intentionally train it to skip certain data. >every time the whole conve
203.
by kgeist 9mo ago
I tried generating code with ChatGPT 5.2, but the results weren't that great: 1) It often overcomplicates things for me. After I refactor its code, it's usually half the size and much more readable. It often adds unnecessary check
204.
by kgeist 10mo ago
>Mafia III was delisted from Steam on May 19th, 2020 when it was replaced with Mafia III: Definitive Edition >Lumines was delisted from Steam on June 22nd, 2018. The delisting coincides with the release of Lumines Remastered just a fe
205.
by kgeist 10mo ago
>It completely breaks the censor's standard playbook of IP enumeration. You can't just block a specific subnet without risking blocking future legitimate allocations At least in Russia, they don't really care about collate
206.
by kgeist 10mo ago
Interesting, zer- seems to be similar to Slavic raz- somewhat. In Russian: davit - press, razdavit - to crush. Siedlung corresponds to Russian selenie "settlement". Zersiedlung appears to correspond to "rasselenie" (morp
207.
by kgeist 10mo ago
>From the home page, I figured it was some text-based game or experiment and closed the page. Same, my first thought was that it's some pentesting game where you're given a VM and your task is to somehow break it. The line &quo
208.
by kgeist 10mo ago
>A world is consistent and has a set of rules that must be followed. Large language models are mostly consistent, but they have mistakes even in grammar too, from time to time. And it's usually called a "hallucination". Ca
209.
by kgeist 10mo ago
>It is not that different from German in this matter. Russian inflection changes the stress. In German it's fixed. Inflectional forms are much more varied in Russian. Colloquial German is much more analytical (past tense is almost a
210.
by kgeist 10mo ago
Just a week ago, I rewrote our RAG pipeline to use structured outputs, and the tests showed no significant difference in quality after a few tweaks (under vLLM). What helped was that we have a pipeline where another LLM automatically scores

More ›