Eric Florenzano
@ericflo.bsky.social
Building assistance! Used to make AR/VR. Before that made mobile apps. Before that made web things. People first, then products, systems, data. he/him
Turns out bf16 sucks, who knew, precision > range github.com/sail-sg/Prec...
October 31, 2025 at 10:14 PM
Turns out bf16 sucks, who knew, precision > range github.com/sail-sg/Prec...
Wow! I thought this number would be way lower, like 1-5%! I thought a whole lot more of y'all were wearing contacts and getting lasik. My model of the world was way off, that's kinda neat actually.
October 25, 2025 at 4:35 AM
Wow! I thought this number would be way lower, like 1-5%! I thought a whole lot more of y'all were wearing contacts and getting lasik. My model of the world was way off, that's kinda neat actually.
Update: it's going away 😢
September 23, 2025 at 11:14 PM
Update: it's going away 😢
Hyperparam sweepin' - first time trying AdEMAMix (runs 0.2.2, 0.2.3, 0.2.4, 0.2.5) - it's good!
September 13, 2025 at 5:40 PM
Hyperparam sweepin' - first time trying AdEMAMix (runs 0.2.2, 0.2.3, 0.2.4, 0.2.5) - it's good!
Started this model training yesterday BTW. Custom 500Mtoken post-training dataset that I distilled from GLM-4.5 Air (single turn), K2-Instruct 0905 (multi-turn and instruction following), and gpt-oss-120b (tool calling) - should be a drop-in replacement for Qwen3-4B-Instruct-2507
September 11, 2025 at 9:19 PM
Started this model training yesterday BTW. Custom 500Mtoken post-training dataset that I distilled from GLM-4.5 Air (single turn), K2-Instruct 0905 (multi-turn and instruction following), and gpt-oss-120b (tool calling) - should be a drop-in replacement for Qwen3-4B-Instruct-2507
For some reason I just thought of this viral "dystopic internet" image from 16 years ago, went and found it, and it's so interesting. I thought it hyperbolic at the time. Now it seems quaint! Sentiment from back then: www.reddit.com/r/pics/comme...
September 9, 2025 at 6:49 AM
For some reason I just thought of this viral "dystopic internet" image from 16 years ago, went and found it, and it's so interesting. I thought it hyperbolic at the time. Now it seems quaint! Sentiment from back then: www.reddit.com/r/pics/comme...
The coding agent is Just Like Me (v2_best_really_final)
August 25, 2025 at 3:05 AM
The coding agent is Just Like Me (v2_best_really_final)
OK gpt-5 generated this pretty good satire about itself! 🤣
August 9, 2025 at 7:35 PM
OK gpt-5 generated this pretty good satire about itself! 🤣
August 7, 2025 at 8:39 PM
This post brought to you by:
July 23, 2025 at 8:22 PM
This post brought to you by:
this but Claude model outage
July 17, 2025 at 9:24 PM
this but Claude model outage
It's never going well when "You're absolutely right!" 😂
July 14, 2025 at 11:56 PM
It's never going well when "You're absolutely right!" 😂
e.g. This is SUCH A GOOD POINT-OF-VIEW about open weights! bigeagle.me/2025/07/kimi...
July 13, 2025 at 4:02 PM
e.g. This is SUCH A GOOD POINT-OF-VIEW about open weights! bigeagle.me/2025/07/kimi...
Why is nobody talking about this? H-Net may actually displace the current popular methods of tokenization. It looks like it quite improves data efficiency. And it feels properly bitter lesson pilled. goombalab.github.io/blog/2025/hn...
July 12, 2025 at 8:08 PM
Why is nobody talking about this? H-Net may actually displace the current popular methods of tokenization. It looks like it quite improves data efficiency. And it feels properly bitter lesson pilled. goombalab.github.io/blog/2025/hn...
This article, from the XFN Wikipedia page, is such a slice of its time. archive.seattletimes.com/archive/2003...
June 24, 2025 at 6:17 PM
This article, from the XFN Wikipedia page, is such a slice of its time. archive.seattletimes.com/archive/2003...
Trying to be optimistic about OpenAI's pending open source model, but finding a lot of reasons to be cynical. I have a feeling the research breakthrough will come with some huge trade-off. Like they found a way for a phone-sized model to benchmark well enough they can go to the poll and say "both!"
June 11, 2025 at 11:17 PM
Trying to be optimistic about OpenAI's pending open source model, but finding a lot of reasons to be cynical. I have a feeling the research breakthrough will come with some huge trade-off. Like they found a way for a phone-sized model to benchmark well enough they can go to the poll and say "both!"
This is such a great explanation! I knew this stuff - and anyone who watches Interstellar or whatever understands at least on some level the interplay between time and space. But this framing really clicks into place in a way that I haven't seen before.
May 31, 2025 at 3:21 PM
This is such a great explanation! I knew this stuff - and anyone who watches Interstellar or whatever understands at least on some level the interplay between time and space. But this framing really clicks into place in a way that I haven't seen before.
Did you just tell me to go F my self? 🤣
May 20, 2025 at 5:53 AM
Did you just tell me to go F my self? 🤣