A lot is said about context you can feed into the LLM but I do think there is still superior power in human context awareness. That kind of ambient collation and organisation of the whole business and its purpose, all the different work going on and how it all relates to eachother. It happens when you isolate business units a bit too much also.
It's not surprising that if you have a hundred separate, isolated contexts working on the same business, that don't cross-talk and have no ability to subconsciously receive and collate, prioritize the thousands of signals we get from our work environment, that you end up shipping lots of incomplete or incompatible work.
Flying drones is an interesting one, I guess you do have to drive the car out to site and set up the drone. But a lot of drone ops are waypointed, automatic flight. I can see a future in which the only thing the operator does is drive the drone van to site, hit the deploy button as the drone pops out the roof, and wait for it to return. Mission set up already by an LLM prompt back at the office.
Super cool and close to my heart, albeit not the Yamanote line for me.
There is an episode of Our Man in Japan with James May where he spends an admittedly short moment with the composer of some(all?) of these melodies. It's a surprisingly thoughtful process, he tries to capture the feeling of the station and area in a short motif. Some of these motifs can contain surprising musicality and complexity, despite being so short.
I have really liked my GL.iNet travel router, also with OpenWRT. I didn't think I would need a travel router but they're pretty handy.
I didn't realise routers like theirs existed, and had been paying through the nose for your standard brands like TPLink and hoping it didn't get popped.
The front page of HN has definitely be sparse on actual software accomplishments, it's been a lot of meta-level chatter on AI. Obviously a very important technology, but also just a tool, a means to an end. I am not seeing much progress in other fields making it to HN regardless of what tool they used. I hope it's happening, and just being drowned out.
The AI can't be held liable so it somewhat defeats the purpose of that kind of paper trail and compliance work. If the threat was losing your livelihood and possibly jail time, would you currently be comfortable being held responsible for AI output?
I think the unlock only happens once though. I think that's where people are misled at the moment, the technology was there but required huge compute and data ingest to show improvement, but we have done that now. What's next for a giant leap is not more compute, and what new data we can provide now pales in comparison to that first ingest.
Because a lot of things in the real world require forward planning, I don't think everything can be just-in-time and tentative/reversible. Some things have to be committed to else you have real consequences and lose your money.
The crowdsource page says it has refillable ink tanks rather than cartridges. Unless they are modifying HP cartridges, which is probably not smart from a legal standpoint.
I think I understand, but I will say that problem solvers are often masquerading as coders. I think they will leave software too. It is exactly the interesting problem solving that goes missing with heavy LLM use. Most business problems are not that challenging, from a problem solving perspective, that's just life. So the interesting part was always the problem solving in the build. I have built things with a huge spectrum of skills and tech, not just software. I learn details fast and have good systems thinking that help me apply that new knowledge.
What LLM usage has changed is that there is no longer a deep dive into domain knowledge, the LLM goes off and does that. Then implementation time comes and again it's just handholding the drunk chatbot inside the codebase until it is done. The whole time my mind is barely engaged. Yes my expertise is required to guide it, I am constantly catching issues and problems it generates, but that's still not engaging the problem solving skillset I developed. It's just leveraging my experience.
I think at that point in my life I quite enjoyed how much he laboured the details. But the book Ready Player One had a few areas where the author just listed pop references, on and on, which reminds me of that.
Something that helps me is just giving myself license to skip stuff. It's usually better I finish a book since I will never come back to it. So I just jump around a few pages if I get bored.
I have certainly noticed my stress skyrocket in this new mode of working. I was used to getting a lot done very quickly, with intense pockets of work followed downtime. Now it feels more like a steady stream of medium stress, and there is no opportunity to stop or drop the thread.
I must admit, if this is the new way of doing software development (eg: not actually programming but working with LLMs) I am not going to stick around for that long. It's not what I fell on love with, it's not what I trained for etc. I may as well do a job I don't enjoy that lets me rest my brain for later.
Somehow this is only the first time I have seen this vector taken advantage of with my own eyes.
I remember thinking it was a stupid idea to embed third party hosted JS back when jquery and prototypejs were the duopoly of javascript. I'm surprised it took me this long to see it.
They are linked, but what I meant was its not a 1:1 relationship. You can get stronger without as much hypertrophy, or you can get bigger without as much strength gain.
The only reason I mention it is because I often see people, especially men, assume getting bigger is the main focus of getting stronger and healthier. But a more holistic approach including stretching, functional strength excercises, cardio and balance training is perhaps more important
I have met people who figured, because they don't excercise they don't wear their body out, so their joints etc. will last longer. Same for injury, no sport no injury, that must be good!
It's also great to have capability to run local models for more brute force tasks. Because you can change the system prompt, you can get local LLMs to do all kinds of high volume tasks without burning through tokens on a hosted model.
Just one example, I needed a bunch of images tagged and organised, with a local vision capable model I could pretty easily set that up and leave it running overnight.
I already had the GPU and memory for gaming, so it was at no cost for me to start running local models. But I feel the long term writing is on the wall, local models will only make more and more sense as they get better and more efficient.
It's not surprising that if you have a hundred separate, isolated contexts working on the same business, that don't cross-talk and have no ability to subconsciously receive and collate, prioritize the thousands of signals we get from our work environment, that you end up shipping lots of incomplete or incompatible work.