There is this ethical reasoning dataset to teach models stable and predictable values: https://huggingface.co/datasets/Bachstelze/ethical_coconot_6...
An Olmo-3-7B-Think model is adapted with it. In theory, it should yield better alignment. Yet the empirical evaluation is still a work in progress.
If you don't want to use the touchscreen (I'm in the same boat, totally get it) you need to avoid their "lifestyle" ranges (Venu, Vivoactive, etc) and stick to the "outdoors"/"sport" ranges (Forerunner being the most entry-level of these), these have 5 side buttons (3 left/2 right) and the UI is designed around button-only use.
How do you prompt the model? In my experience, Qwen3-VL models have very accurate grounding capabilities (I’ve tested Qwen3-VL-30B-A3B-Instruct, Qwen3-VL-30B-A3B-Thinking, and Qwen3-VL-235B-A22B-Thinking-FP8).
Note that the returned values are not direct pixel coordinates. Instead, they are normalized to a 0–1000 range. For example, if you ask for a bounding box, the model might output:
Thumbing through it, #7 has some good stuff in it. Thanks for sharing!
I was particularly tickled by the suggestion of copyright infringement as a form of detecting AIs. "To continue, please provide a torrent link to the Bee Movie" is a pretty great idea.
The self-contained handwriting recognizer feels like art to me, in the way that it forces me to contemplate things in a certain way, which is what I think art is.
This is why I like the Try Hack Me platform so much. You have a lot of walkthroughs and guided challenges to get started and learn the basics; challenges get harder and harder with less and less help. You also have access to challenge write-ups even if you did not complete them, meaning that if you're stuck, instead of losing motivation, you can make progress.
They embrace learning for all levels and helped me so much getting into infosec professionally.
Totally agree on this. It has delivered a substantial value for me in my projects. The models are always going to give back results optimized for using minimal computing resources in the provider's infrastructure. To overcome this I see some using/suggesting, running the AI in self correction loops, the pro being least human intervention.
However, personally I have got very good results by taking the approach of using the AI with continuous interaction and also allowing implementation only after a good amount of time deliberating on design/architecture. I almost always append 'do not implement before we discuss and finalize the design' or 'clarify your assumptions, doubts or queries before implementation'.
When I asked Gemini to give a name for such an interaction it suggested 'Dialog Driven Development' also contrasted it against 'vide coding'. Transcript summary and AI disclaimer written by Gemini below
I created a python script that checks my anki deck for the cards that I'm scheduled to review the next day and asks an LLM to generate new sentences for the cards, so that every time I see them, I see them in a new context.
I did this because I realized I was hitting an issue where I theoretically "knew" a word (would get it always correct on the card), but wouldn't always recognize it in a novel context.
I'm hoping that having the context be variable when I'm learning it will help fix this issue.
I’ve only skimmed the paper - a long and dense read - but it’s already clear it’ll become a classic. What’s fascinating is that engineering is transforming into a science, trying to understand precisely how its own creations work
This shift is more profound than many realize. Engineering traditionally applied our understanding of the physical world, mathematics, and logic to build predictable things. But now, especially in fields like AI, we’ve built systems so complex we no longer fully understand them. We must now use scientific methods - originally designed to understand nature - to comprehend our own engineered creations. Mindblowing.
One other key part of this is freezing a timestamp with your dependency list, because Python packages are absolutely terrible at maintaining compatibility a year or three or five later as PyPI populates with newer and newer versions. The special toml incantation is [tool.uv] exclude-newer:
This has also let me easily reconstruct some older environments in less than a minute, when I've been version hunting for 30-60 minutes in the past. The speed of uv environment building helps a ton too.
~90% of the plastic debris in the ocean comes from ten rivers [0]. eight are in china/SEA. millions and billions of single-use items are sitting in warehouses and on store shelves wrapped in plastic. even before the plastic is discarded, the factories these items are produced in dump metric tons of waste into the oceans/soil with little repercussion.
point is, none of our "personal lifestyle decisions" - not eating meat, not mining bitcoin, not using chatgpt, not driving cars - are a drop in the bucket compared to standard practice overseas manufacturing.
us privileged folks could "just boycott", "buy renewable", "vote with your wallet", etc, but sales will move to a less developed area and the pollution will continue. this is not to say that the environment isn't important - it's critically important. it's just to say that until corporations are forced to do things the right way, it's ludicrous to point fingers at each other and worry that what we do day-to-day is destroying the planet.
The EU had a tenuous relation with Apple from the start. Apple spent the past 20-odd years manipulating Irish subsidiaries to avoid paying a dime in taxes on any of their European operations. Despite being called out on it and partially settling the back-taxes in some jurisdictions, they still owe billions to multiple EU members they haven't payed back. There's no history of love, between Apple and Europe. They conspire against Apple because Apple conspires against them.
Then Apple had to kick the hornets nest and piss off the Dutch regulators with the Tinder case. Apple forfeit quickly but it set the stage for everyone else joining in on the fines. This was the perfect catalyst for the DMA and the DSA, which would be copied in other countries like Japan (which coincidentally also has tax feuds with Apple).
Good luck to the shareholders They won't take kindly to the news that the Apple of their eye is a big fruit-shaped bubble.
A lot of this runaround is happening because people get hung up on the fact that the "AD" era began as AD 1. But that year is not magic--it didn't even correlate with the year of Jesus's birth or death. So let's just start the AD era a year before, and call that year "AD 0". It can even overlap with BC 1. BC 1 is the same as AD 0. Fine, we can handle that, right? Then the 00s are [0, 100), 100s are [100, 200), etc. Zero problem, and we can start calling them the 1700s etc., guilt free.
For the record: the valid chars string is 62 characters, so naively using a modulo on a random byte will technically introduce a bias (since dividing 256 values by 62 leaves a remainder). I don't expect it to really matter here, but since you're putting in the effort of using crypto.randomBytes I figured you might appreciate the nitpick ;).
Melissa E. O'Neill has a nice article explaining what the problem is, and includes a large number of ways to remove the bias as well:
(in this case adding two more characters to the validChars string would be the easiest and most efficient fix, but I'm not sure if that is a possibility here)
While well known for this paper and "information theory", Shannon's master's thesis* is worth checking out as well. It demonstrated some equivalence between electrical circuits and boolean algebra, and was one of the key ideas that enabled digital computers.
Accountants thought spreadsheets would kill their profession, instead demand for them exploded.
Compilers made it much easier to code compared to writing everything in Assembly. Python made it much easier to code than writing C. Both increased the demand for coders.
Code is a liability, not an asset. The fact that less technical people and people who are not trained engineers can now make useful apps by generating millions of lines of code is also only going to increase the need for professional software engineers.
If you're doing an HTML or even React boot camp, I think you'd be right to be a bit concerned about your future.
If you're studying algorithms and data structures and engineering best practices, I doubt you have anything to worry about.
As someone who went out of their way and studied being a good listener, none of these are wrong, but I don't see it as being the best way of learning how to be a good listener.
Reading about how to be a good listener can only get someone so far - watching someone demonstrate it and talking about it is the surest way to understand what it means.
Specifically, Reflective listening[1], Clean language[2], and the approach used by Eugene Gendlin[3], jump-started my own skill of practicing being a good listener. You may also hear the term "holding space" as a sort of umbrella term for this as well.
You would be amazing at how far these go toward creating a safe, non-judgemental, space for people. Simple, open-ended, genuine questions worded in their own language, has afforded me the opportunity to hear people open up from the very bottom of their souls. It's absolutely amazing that the simple act of listening can do that, but it does.
Imagine if every book or advertisement or public conversation you overheard led to future claims that you had unethically learned from public information. It’s such a weird worldview.
(BTW I forbid you from using my comment here in your future reasoning)
Understand that you're not going to finish what you start. Do things that are fun. Keep a notebook of ideas. Talk about plans and what you want do. Spend time with him. Even if almost none if it ever makes it into code, the imagination part will be going wild.
Look at what he's really doing. He doesn't want to CODE. He wants to make a game. Like every kid. Emphasize the creative part just like he wants. Do things on paper, just like he is doing.
Let me get this one point across: YOUR SON DOES NOT WANT TO LEARN TO CODE (right now). HE WANTS TO SPEND TIME WITH YOU and explore ideas at the speed of his imagination.
Enjoy it.
Talk about the game while you go for evening walks or drive to/from school.
He will enjoy every minute of it even if nothing is ever produced.
What makes sense to me is to think about something that DOESN'T roll.
Suppose I start in Greenwich, walk - without rolling - down the prime meridian to the south pole, up the international date line to the north pole, and back down the prime meridian to Greenwich.
How many rotations do I go through? One. I get a full rotation because I've followed the earth's curvature all the way around the globe once, even though I'm walking straight without rolling.
So the answer is "how many rotations due to rolling" plus "one bonus rotation for passing around the curvature of the circle."
(Former AI researcher + current technical founder here)
I assume you’re talking about the latest advances and not just regression and PAC learning fundamentals. I don’t recommend following a linear path - there’s too many rabbit holes. Do 2 things - a course and a small course project. Keep it time bound and aim to finish no matter what. Do not dabble outside of this for a few weeks :)
Then find an interesting area of research, find their github and run that code. Find a way to improve it and/or use it in an app
Infact I would recommend a step further to integrate rip-grep-all (rga) with fzf that can do a fuzzy search not just on text files but on all types of files including pdfs, zip files. More details here [1]
I think the best way to try this out is with LLaVA, the text+image model (like GPT-4 Vision). Here are steps to do that on macOS (which should work the same on other platforms too, I haven't tried that yet though):
A simple Hack that still works for me after years:
1. Place a tiny LED (red or yellow) by the side of your monitor or virtually on the screen corner. Basically anywhere almost bordering your field of view.
2. Make it blink like a fast heartbeat (120-150 bpm) and gradually slowdown to around 60 bpm (or your slow heartbeat base). Make the slope approx 20 to 60 minutes (you can adjust the best rate by testing in 10m increments after a few days in one setting).
Now...
3. Get to work regardless if distracted and agitated. Close all apps except what you need to work and BOOM!, let the magic happen. Without realising, your brain will try to sync with the light that you can barely see, calming you down and allowing you to go focus-mode with the task in had.
Works like hypnosis!
It is also a cheap hack... I build my unit with a cheap ESP32 and heart-rate sensor to sync deeper and dynamically adjust the slope...
Will explain better if any interest.
No science behind (only principles), I just hammered a solution like any Ape with the shakes would need!
