Joel Martínez

493 posts

Joel Martínez banner
Joel Martínez

Joel Martínez

@_j0x0j_

Building Tools 🌎🛰️ for⚡️Grid Operators | Prev Co-Founder and CTO @getraincoat | Organizer of 🇵🇷 Cloud Native San Juan

San Juan Katılım Ocak 2011
1.2K Takip Edilen304 Takipçiler
Joel Martínez
Joel Martínez@_j0x0j_·
Today’s office. Captured eucalyptus, mango, mahagony, african tulip and bamboo. All in the distribution strike zone. The canopy height model proved to be accurate.
English
0
0
3
38
Hubert Thieblot
Hubert Thieblot@hthieblot·
Founders: What's the one thing you do to relax? For me, it was always playing video games 🎮
English
253
3
289
23.8K
Joel Martínez
Joel Martínez@_j0x0j_·
data labeling and fine tuning hours, my tools have gotten better but i need to go way faster
English
0
0
0
110
Joel Martínez
Joel Martínez@_j0x0j_·
@patrickc @tobi my openclaw has been running ml research experiments for a few weeks now, that's why i gave it a NVDIA GPU for CUDA support. The bottleneck now is me, labeling data.
English
0
0
0
1.1K
Patrick Collison
Patrick Collison@patrickc·
@tobi Two weekends ago, I asked Claude to train a weather forecasting model on 6+ years of historical data I had. After training initial model, I had it generate hypotheses for how to improve the architecture, test them, and then integrate learnings. Worked great. RSI is totally here.
English
25
27
977
76.2K
tobi lutke
tobi lutke@tobi·
OK this thing is totally insane. Before going to bed I... * used try to make a new qmdresearcher directory * told my pi to read this github repo and make a version of that for the qmd query-expansion model with the goal of highest quality score and speed. Get training data from tobi/qmd github. * woke up to +19% score on a 0.8b model (higher than previous 1.6b) after 8 hours and 37 experiments. I'm not a ML researcher of course. I'm sure way more sophisticated stuff is being done by real researchers. But its mesmerizing to just read it reasoning its way through the experiments. I learned more from that than months of following ml researchers. I just asked it to also make a new reranker and its already got higher base than the previous one. Incredible.
Andrej Karpathy@karpathy

I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then: - the human iterates on the prompt (.md) - the AI agent iterates on the training code (.py) The goal is to engineer your agents to make the fastest research progress indefinitely and without any of your own involvement. In the image, every dot is a complete LLM training run that lasts exactly 5 minutes. The agent works in an autonomous loop on a git feature branch and accumulates git commits to the training script as it finds better settings (of lower validation loss by the end) of the neural network architecture, the optimizer, all the hyperparameters, etc. You can imagine comparing the research progress of different prompts, different agents, etc. github.com/karpathy/autor… Part code, part sci-fi, and a pinch of psychosis :)

English
120
238
4.8K
797.1K
Joel Martínez
Joel Martínez@_j0x0j_·
I was not authorized to fly in Tyrannus-controlled airspace
Joel Martínez tweet media
English
0
0
1
172
Joel Martínez
Joel Martínez@_j0x0j_·
cloning the openclaw workspace repo from my headless device to my laptop has been really helpful
English
0
0
0
155
Joel Martínez
Joel Martínez@_j0x0j_·
@manofhisword It's more about understanding the threat model at each step and the tradeoffs you are willing to make for gaining capabilities.
English
0
0
0
55
Digital Oil Hunter
Digital Oil Hunter@manofhisword·
@_j0x0j_ love the incremental trust model. what signals are you looking for before expanding permissions? curious how you're thinking about the trust boundary between agent and host
English
1
0
0
78
Joel Martínez
Joel Martínez@_j0x0j_·
running openclaw on a jetson orin nano, really impressed so far. i started with a sandboxed setup and slowly gave it more permissions and capabilities as i understood how it works. it runs a docker container and mounts a workspace from the host, read/writes can only happen there if you want. it is also isolated from the host unless you want the main agent to use your host user's permissions. i also installed tailscale and added the device to my network so that i can reach it from anywhere. i would recommend to use it over the default web chat before connecting any external messaging apps. also, when i want to add a new capability i just ask it how. i've been sharing research docs, tabular data and tiffs. it is performing suprisingly well and contributing valuable geospatial work.
Joel Martínez tweet media
English
1
0
2
448
kitze
kitze@thekitze·
i'm using both opus 4.6 and codex 5.3 in parallel and i fkn love them both and feel like i took some limitless pill or some shit lmao IT'S FEBRUARY 6TH !!!!!
GIF
English
29
9
254
10K
Joel Martínez
Joel Martínez@_j0x0j_·
measuring conductor to vegetation clearance in blender. i now have a complete FOSS workflow which mainly costs me a few dollars in google colab, takes around 80 min to process 10 million points using an H100. i will later share the notebook.
English
0
0
2
211
Joel Martínez
Joel Martínez@_j0x0j_·
I was inspecting 3DEP LiDAR tiles and suddenly i found a gold mine; catenary data that i can fuse into my photogrammetry models, which suffer from sparse points around conductors.
Joel Martínez tweet media
English
2
6
119
7.6K
Nikita Bier
Nikita Bier@nikitabier·
Over the last few months, we scoured the world for the top posters in every niche & country We've compiled them into a new tool called Starterpacks: to help new users find the best accounts—big or small—for their interests ⬇️ Reply below with a topic you're most interested in We'll be rolling out to everyone in the coming weeks.
English
8.5K
591
10K
3.2M
Joel Martínez
Joel Martínez@_j0x0j_·
blender's controls are not intuitive (for me). struggling to set up a smooth workflow to analyze my point clouds, will have to watch more tutorials.
English
0
0
0
167
Joel Martínez retweetledi
Simon Willison
Simon Willison@simonw·
I used Claude Code to reverse-engineer the Claude macOS Electron app and had Cowork dig around in its own environment - now I've got a good idea of how the sandbox works It's an Ubuntu VM using Apple's Virtualization framework, details here: gist.github.com/simonw/35732f1…
English
19
54
718
206.5K
Joel Martínez
Joel Martínez@_j0x0j_·
got my CAGE code today, just a few hours before a deadline to submit a paper... made it just in time to be the prime.
English
0
0
0
139
Joel Martínez retweetledi
Kanahiro Iguchi
Kanahiro Iguchi@kanahiro_iguchi·
書き初め:Sentinel-2 Viewer Sentinel-2の衛星画像を検索して地図上で表示するWebアプリケーション sentinel.spatialty.io
日本語
2
43
302
49K
Joel Martínez retweetledi
Qiusheng Wu
Qiusheng Wu@giswqs·
Compare Layers Visually in QGIS — New Leafmap Plugin Demo! Discover the new QGIS plugin for Leafmap that makes comparing geospatial layers easier than ever! Whether you're working with raster or vector data, this tool brings interactive transparency controls and a layer swipe feature right into your QGIS environment. GitHub repo: github.com/opengeos/qgis-… Video tutorial: youtu.be/glBgnyS8IDY #QGIS #geospatial #opensource #datascience
YouTube video
YouTube
English
1
27
203
9.2K
Joel Martínez
Joel Martínez@_j0x0j_·
Big week, everything is lining up.
English
0
0
0
160