dev exp. | ex @weaviate_io @strapijs | indulging my creative urges | dj mytrax.link/mdltso | saut dans la vide | the man with the axe
Joined November 2015
- Tweets 20,542
- Following 951
- Followers 2,217
- Likes 24,097
1,787 Photos and videos
Pinned Tweet
26 Feb 2021
"because I'm doing this for the thrill of it, killin' it, never not chasin' a million things I want"
malgamves.dev/
1
14
Jun 12
my childhood 😭😭😭
Unlike Meta, these bad boys never go down.
ALT Encyclopedia Britannica volumes on a shelf
40
Jun 11
benchmark experts, what's the best way to build a bench?
Jun 11
We need more niche benches.
We need ios-bench.
We need ts-bench.
We need baseball-bench.
We need yt-thumbnail-bench.
We need way more creativity in how we measure what models can do.
21
Jun 11
how can i do this for friends
Jun 10
i hooked my whoop to my work calendar to find which coworker gives me the most stress 🚨
thanks to fable, I reverse engineered whoop to pull per minute heart rate. nd matched spikes with cal events and attendees
I now have a leaderboard and I think about it daily.
few info masked for obvious reasons ;)
74
Jun 4
relevant french langauge exam prep video of the morning
"Tokenmaxxing" : le présentéisme, version IA 🤳
"C’est considéré comme un marqueur d’implication."
🎙️ @AnthonyMorel
1
24
daniel phiri retweeted
Jun 3
Meet Gemma 4 12B!
A unified, encoder-free multimodal model designed to bring high-performance intelligence directly to your laptop, and released under an Apache 2.0 license.
Bridging the gap between edge efficiency and advanced reasoning. Here is what’s new with Gemma 4 12B: 👇
404
1,789
12,365
3,175,976
Jun 3
ok i'm officially obsessed!
1
47
daniel phiri retweeted
Jun 1
another week at the prompt factory.
23
193
2,557
70,801
Jun 1
the pressure to be dressed to the nines when you get good seats at roland garros 😭
46
May 27
people using lm studio for local models, what main model <> draft model combinations have you been using for speculative decoding?
2
75
// Memory as a Model //
The paper augments any LLM with a separate trained memory model that stores, retrieves, and integrates facts on its behalf.
It decouples memory updates from base-model weight updates. It achieves continual-learning robustness without catastrophic forgetting, which is a property that RAG fails to deliver.
A vector store is a database with a learned encoder bolted on. MeMo is a learned subsystem with explicit interfaces. That distinction matters, as agents need to be able to ingest fresh knowledge weekly without retraining or vector-DB churn.
At its core, the position here is that memory in agents should be modular, learned, and gated, not a context-window hack.
Paper: arxiv.org/abs/2605.15156
Learn to build effective AI agents in our academy: academy.dair.ai/
23
112
599
66,342
daniel phiri retweeted
May 24
It was cycling through Paris going back to where I was staying after a day of walking around with @malgamves . I would go back to that day any time.
Walking alone through a foreign city at night, realizing how far you’ve come. Top 3 peak moments ever.
1
2
3
484
daniel phiri retweeted
May 23
The @enhanced_games open in Las Vegas this weekend.
Before the starting gun — @analog_kyle sat down with CEO @max_enhanced to ask the question everyone’s thinking: if 43% of Olympians are already doping, who’s really playing by the rules?
Full episode on YouTube — link in bio. Follow the games live on @enhanced_games
1
4
5
3,738
daniel phiri retweeted
May 17
not one mention of suffering in carrick’s speech
15
786
5,980
54,495
daniel phiri retweeted
May 16
Getting opportunities that scare the shit out of you >>>
136
9,449
46,966
633,708
daniel phiri retweeted
May 17
Underrated life advice: Make yourself easy to root for. Be kind. Be reliable. Celebrate other people’s wins. Work hard without complaining. Carry good energy into rooms. You'll be shocked by how many doors open for you by making life better for others.
172
3,612
23,564
503,769
