HN user
One thing I’ve noticed is that as my context grows, often performance degrades. So how are you battling your agents being exposed to too many descriptions? I how this works in curated agents where you’re tending it like a garden, but not when we’re looking for organic discovery of how to accomplish a task. It feels like order matters a lot there.
I use this feature DAILY at work, you built something great here. I tend to write in md locally, this makes sharing the work with others easy. Especially to those less plaintext inclined.
One thing I did notice, I can't seem to find a way to set a default codeblock font format. The default font option isn't totally monospaced and so some ascii art looks weird :/
I don't think that has anything to do with your contribution though.
THANK YOU!
Yes that is a misconception: https://data.worldobesity.org/country/mexico-139/#data_preva...
1. The Pragmatic Programmer.
2. Writing Tools: 50 Essential Strategies for Every Writer.
3. Diffusion of Innovations.
4. The Little Book of Common Sense Investing.
5. Implementing Lean Software Development: From Concept to Cash.
6. Writing to Learn.
7. On Writing Well.
8. Pragmatic Thinking and Learning: Refactor Your Wetware.
9. Thinking In Systems: A Primer.
10. The Startup Owner's Manual: The Step-By-Step Guide for Building a Great Company.
+1 - Shared with my wife, who is also a DPT(Southeastern US) and she agreed with all of these points.
I've finally made it to the $500/month mark! I built https://wolftickets.ai , it is a collection of AI predictions for upcoming UFC fights. The predictions for the future events are private but all past results are public.
I get to keep iterating on new models, new approaches for using gen-ai tech to write better analysis of fights and fighters, along with exploring stats and tendencies that matter.
A good example of the writeups: https://wolftickets.ai/events/ufc-310-pantoja-vs-asakura/139...
This project isn't yet allowing me to retire but I'm passionate about the AIML space and combat sports, I get to explore whatever ideas I find interesting, and get a ton of feedback and ideas from members without having to do advertising.
Chiming in with many others here. When I was 3 my mom would take me into her office on occasional weekends and she managed to get me to memorize enough to open The Print Shop. From there I could print all sorts of dinosaurs and interesting things to a dot matrix printer, and then color as much as I wanted. It connected tech to me as a tool to reach for at a very young age. I definitely would have had a very different route into tech later if I even had one. Thank you all!
I assume via Jupyter Notebook or Lab ( not VS Code running it )
You wouldn’t really be able to tell. Figma is often just a tool for working out designs and building up a design system for your site. That later gets translated to your front end by your devs. Figma itself isn’t a styling library like tailwind, etc.
Protect AI | Product Designers, Software Engineers & Solution Architects | Seattle, Washington | Full Time | $140-$200k approx | https://protectai.com
Protect AI is shaping, defining, and innovating a new category within cybersecurity around the risk and security of AI/ML. Our ML Security Platform enables customers to see, know, and manage security risks to defend against unique AI security threats, and embrace MLSecOps for a safer AI-powered world. This includes a broad set of capabilities including AI supply chain security, Auditable Bill of Materials for AI, ML model scanning, signing, attestation and LLM Security. We are expanding our team in Seattle and have a number of senior roles open:
* Senior Product Designer
* Senior Product Manager
* Senior Software Engineer: Backend
* Senior Software Engineer: Frontend
* Senior Software Engineer: Infrastructure
* Senior Solutions Architect: AI/ML
* Senior Solutions Architect: Infrastructure
Feel free to send me your resume - chris (at) protectai.com - or you can apply directly here: https://protectai.com/careers
Full disclosure I am head of product at Protect AI. To make this easier for everyone we have an open source tool (friendly licensing) called ModelScan https://github.com/protectai/modelscan/tree/main I wouldn't be shocked if they are using this under the hood, but all the best if they are! For a bit more info on this type of attack: https://protectai.com/blog/announcing-modelscan
If you happen to have a GitHub space for storing things you can upload release artifacts there for folks to download.
Your problem isn't because you are a non native speaker. I'm a native speaker and struggled to understand their point. They need to write in a more concise manner.
Read all of those but wanted to chime in that when I was in college the rise of TR inspired me to get my life in order. Not that he was an ideal but it was a lens to see a life that was focused and what he could tackle. Stellar writing.
They hate us because they ain’t us
Just wanted to share, the charts and gifs are exceptionally well done. Informative, concise, and easy to read.
[I work at Protect AI] You're spot on for data being the jugular, interestingly with exploits like this as an attacker you could quickly go for attacking model content but also have credentials that would grant you access to data in many cases.
These tools can serve as the first opening but a sizable one when looking to attack an enterprise more broadly.
[I work at Protect AI] - The goal here was initially relatively common tooling around MLOps/Data Science work. All ears here if you have some ideas for other projects to explore.
Absolutely, though that isn't strictly what we're talking about here.
In this case, models themselves are fundamentally files. These files can have malicious code embedded into them that is executed when the model is loaded for further training or inference. When executed it isn't obvious to the user at all. It's a very nasty potential vector.
I wrote a blog about it here: https://protectai.com/blog/announcing-modelscan
Disclosure I work for the company that released this: https://github.com/protectai/modelscan but we do have a tool to support scanning many models for this kind of problem.
That said you should be using something like safe-tensors.
I lead product at Protect AI and we just released ModelScan. It is open source project that scans models to determine if they contain unsafe code. It is the first model scanning tool to support multiple model formats. ModelScan currently supports: H5, Pickle, and SavedModel formats. This protects you when using PyTorch, TensorFlow, Keras, Sklearn, XGBoost, with more on the way.
On the other hand I can’t fathom a better workout partner than my Apple Watch. A few seconds from racking the weights or whatever I can fire up my rest timer with 1 press and mute the vibration with another press when it’s time to get back to lifting. This works great when you want to do some interval of rest paired later with intervals of activity. Having favorited a few timers I can reach them within 1 to 2 presses. I haven’t found a similar watch with a range of timers that are as easy to use.
Protect AI | Multiple Roles | Full-time | Seattle, Dallas, Raleigh | https://protectai.com/careers
Protect AI is working to build a safer world through improving the security of AI and ML workloads. We’re funding ML security researchers, finding unique exploits, providing tools to detect and mitigate risks found in operating ML pipelines. We’re working to put MLSecOps at the heart of ML practices everywhere.
We have open positions for SDEs, Cybersecurity Engineers, and Red Team Cyber Security Engineers for AI/ML, and AI/ML Engineers as well.
If you’re curious to learn more, apply on our site or DM me for more information.
Not the exact same but I just built a machine for machine learning(tabular work mostly) using a Ryzen 9 5950X(16 core) and a 3060 GPU and the performance is stellar with Linux.
I think they mean including a 4090 in a build with either CPU choice would produce a more performant computer.
There’s actually a huge volume of research on spatial components mapping to memory. Check out the book: Moonwalking with Einstein for proper citations.
LG 32UN880, it’s 32” and 4k, works great!
Absolutely! This has taken my 200 initial stake to bit over 5200 now.
https://wolftickets.ai/ , right now 1600ish MRR, started as a way to really dig into machine learning on a problem I found super interesting. Has evolved into me learning more and more about making products. Lots of work to do still!
What?!