antirez
@antirez.bsky.social
Reproducible bugs are candies 🍭🍬
Required remark about Fil-C
November 7, 2025 at 7:25 AM
Required remark about Fil-C
Hi friends, this summer I published my latest sci-fi short story with Urania-Mondadori. I attempted to translate it in English, for the potential English speaking readers here, and I believe the result is truthful to the original. This is the ePub file.
antirez.com/sincro-en.epub
antirez.com/sincro-en.epub
October 21, 2025 at 7:50 AM
Hi friends, this summer I published my latest sci-fi short story with Urania-Mondadori. I attempted to translate it in English, for the potential English speaking readers here, and I believe the result is truthful to the original. This is the ePub file.
antirez.com/sincro-en.epub
antirez.com/sincro-en.epub
Why I believe there are a lot of misleading papers about LLM reasoning.
October 10, 2025 at 6:08 AM
Why I believe there are a lot of misleading papers about LLM reasoning.
ADV: my C course on YouTube is free and can be watched without problems with subtitles or dubbed audio by non Italian speakers :)
September 1, 2025 at 11:34 AM
ADV: my C course on YouTube is free and can be watched without problems with subtitles or dubbed audio by non Italian speakers :)
Playing with adding LLMs embeddings into Redis vector sets, to explore relationships and differences with embedding words models like Word2Vec. Will make a YouTube video and/or blog post.
August 25, 2025 at 2:37 PM
Playing with adding LLMs embeddings into Redis vector sets, to explore relationships and differences with embedding words models like Word2Vec. Will make a YouTube video and/or blog post.
That's the best explanation I got of a technique I developed during Covid for Z80 (Spectrum) 3D programming.
August 8, 2025 at 9:57 AM
That's the best explanation I got of a technique I developed during Covid for Z80 (Spectrum) 3D programming.
Quite happy with vector sets VRANDMEMBER distribution. HNSWs help a lot, with their structure.
July 11, 2025 at 11:53 AM
Quite happy with vector sets VRANDMEMBER distribution. HNSWs help a lot, with their structure.
Don't ask me why sometimes I need to reach such a point.
July 10, 2025 at 5:22 PM
Don't ask me why sometimes I need to reach such a point.
Preparing the YouTube video series on Redis Vector Sets I just discovered a fancy mxbai-large embedding model "issue". It over-expresses a specific component in every output I tested. This may be some very generic/ubiquitous feature the model needed to model based on training.
July 7, 2025 at 12:48 PM
Preparing the YouTube video series on Redis Vector Sets I just discovered a fancy mxbai-large embedding model "issue". It over-expresses a specific component in every output I tested. This may be some very generic/ubiquitous feature the model needed to model based on training.
About the recent XOR article in Hacker News.
July 3, 2025 at 10:34 AM
About the recent XOR article in Hacker News.
Are you from Malta? My uncle moved there years ago and now he is reinvented himself restoring wood doors, windows, and so forth. If you need his services please drop me a DM or an email and I'll reply with his phone number.
June 7, 2025 at 10:06 AM
Are you from Malta? My uncle moved there years ago and now he is reinvented himself restoring wood doors, windows, and so forth. If you need his services please drop me a DM or an email and I'll reply with his phone number.
I understand that in Germany / Austria those waters are very popular. I tried to understand what's the point but then realized it and I'm providing with an improved version here.
From a discussion with my brother living in Vienna.
From a discussion with my brother living in Vienna.
May 27, 2025 at 12:03 PM
I understand that in Germany / Austria those waters are very popular. I tried to understand what's the point but then realized it and I'm providing with an improved version here.
From a discussion with my brother living in Vienna.
From a discussion with my brother living in Vienna.
That's a very nice comment I received about "Kilo", a small side project that took one or two days but resulted in very large effects.
May 27, 2025 at 11:33 AM
That's a very nice comment I received about "Kilo", a small side project that took one or two days but resulted in very large effects.
Worth posting here as well.
May 15, 2025 at 4:55 PM
Worth posting here as well.
Today with Vincenzo Agrillo I recorded an interesting video where he shows how he is using vector arithmetic with Redis vector sets for image search, for the Church. Soon on my channel, a few days to edit the video.
May 8, 2025 at 9:17 PM
Today with Vincenzo Agrillo I recorded an interesting video where he shows how he is using vector arithmetic with Redis vector sets for image search, for the Church. Soon on my channel, a few days to edit the video.
As long as there is X axis symmetry, all is fine :D Reason why solar does not produce in excess when possible is that I'm still not connected to the grid to export energy, it takes a few weeks here in Italy for the paperwork and activation.
May 3, 2025 at 11:40 AM
As long as there is X axis symmetry, all is fine :D Reason why solar does not produce in excess when possible is that I'm still not connected to the grid to export energy, it takes a few weeks here in Italy for the paperwork and activation.
In case you have an inverter that looks like this, regardless of the actual brand you see in the case, you can talk with it via UDP using the Python library "goodwe" (GoodWe is the actual Chinese producer). It works great.
May 3, 2025 at 10:14 AM
In case you have an inverter that looks like this, regardless of the actual brand you see in the case, you can talk with it via UDP using the Python library "goodwe" (GoodWe is the actual Chinese producer). It works great.
3 years ago a HN user fingerprinted the whole community using the Burrows method. I did it again, to reproduce the work done (statically) more dynamically, using Vector Sets. I also visualized the users vectors, like this one (for the PG and few more)
Blog: antirez.com/news/150
Blog: antirez.com/news/150
April 16, 2025 at 2:02 PM
3 years ago a HN user fingerprinted the whole community using the Burrows method. I did it again, to reproduce the work done (statically) more dynamically, using Vector Sets. I also visualized the users vectors, like this one (for the PG and few more)
Blog: antirez.com/news/150
Blog: antirez.com/news/150
In this example I added the same usernames processing a different set of comments (A, B), and added [b] to the usernames of the second set. Of course patio11 is still similar to patio11 of the future :D But also similar to that users.
April 15, 2025 at 12:56 PM
In this example I added the same usernames processing a different set of comments (A, B), and added [b] to the usernames of the second set. Of course patio11 is still similar to patio11 of the future :D But also similar to that users.
Three years ago an HN user posted a very cool web site where users with similar writing style were detected, using words frequency in the comments made within the site. Since I started working with Vector Sets, I thought I would love to replicate the work... blog + video in a few days:
April 15, 2025 at 12:31 PM
Three years ago an HN user posted a very cool web site where users with similar writing style were detected, using words frequency in the comments made within the site. Since I started working with Vector Sets, I thought I would love to replicate the work... blog + video in a few days:
Better uses than the Studio Ghibli style transfer? Of course.
April 2, 2025 at 11:27 AM
Better uses than the Studio Ghibli style transfer? Of course.
Sure! At the end of the exchange in the images it produced the final image.
March 30, 2025 at 8:54 AM
Sure! At the end of the exchange in the images it produced the final image.
It is incredible what the new ChatGPT autoregressive image-gen can do. Yet with complicated tasks it still has issues, but it's a matter of time, of course, not even much time.
March 29, 2025 at 10:21 PM
It is incredible what the new ChatGPT autoregressive image-gen can do. Yet with complicated tasks it still has issues, but it's a matter of time, of course, not even much time.
I just received this personal vs company time accounting device :D
March 24, 2025 at 3:05 PM
I just received this personal vs company time accounting device :D
With M=32, int8 quantization (a lot less extreme than binary quantization) reaches very high (94.5%) level of recall even at EF=250 (very low). Sometimes systems stress *very* high EF values and M values just to show a 99% alike percentage: not worth it, IMHO: slow and memory inefficient.
March 15, 2025 at 12:14 PM
With M=32, int8 quantization (a lot less extreme than binary quantization) reaches very high (94.5%) level of recall even at EF=250 (very low). Sometimes systems stress *very* high EF values and M values just to show a 99% alike percentage: not worth it, IMHO: slow and memory inefficient.