That’s our new reality. Some people seem not to not grasp that all those AIs are just mathematical models producing the next most statistically likely token. It doesn’t feel anything, nor does it care about what it does. The difference between test and production environment is just a word. That, in contrast to a human who would typically have a voice in the back of his head “this is production DB, I need to be careful”.
I don't see any meaningful performance improvements in those paid models anymore.
They all roughly produce junior developer-level code, continue to have mental breakdowns in their “thinking” stage, occasionally hallucinate things, delete pieces of code/docs they don’t understand or don’t like, use 1.5 times the necessary words to explain things when generating docs and so on.
I'm now testing "avoid sycophancy, keep details short and focus on the facts" in my AGENTS.md files.
I know of a publicly traded company which in its early years was built on beer. Literally. 3 guys in a co-working space in Cambridge, MA. Beer fueled their progress. 15 years later the software is still the backbone of the org.
I wanted home assistant compatible plant watering solution that works on a solar panel and does not require being connected to the water line and is Zigbee compatible. Unfortunately, I could not find any. So I did a DIY solution: a big barrel which I manually fill with water, a 12V pump (usually sold for camper vans), some rechargeable batteries, 10W solar panel, a solar charging controller, and Tuya ZG-2002-RF switch.
I dabbled in hydroponics for some years. Due to my inability to get my Rasberry Pi and Arduino working I ended up using a 12V pump and one of those cheap $10 electric timers on Aliexpress. I estimated how much time it took for my hydroponics system to drain and that's what I set on the timer. Other people I followed had all sorts of sensors rigged up, which I would have done if I had the time and skill but I failed, so in the end it was just the timer. I too had single solar panel and battery and the system worked for over 7 years with no issues. I just replaced the pump once or twice.
An art student of mine once needed a way to electrically control precise small amounts of water. We solved that using:
1. Water tank and gravity
2. Medical IV flow regulator¹
3. Servo hooked up to that IV flow regulator via a 3D-printed part
It worked very well. In medical applications off must be really off, so it was also quite safe in that regard as well. Her 3D-printed part had a little bit too much flex in it, but in principle this works quite well. If it is really, really safety critical I would still recommend a mechanical fallback that protects in case of power loss or when the servo fails open (e.g. bending the hose with the force of a spring if electricity is gone).
Your suggestions should be fine for hardware failure but I'd be more concerned about software failure: what if a bug in your software makes it unresponsive and stuck in the state with the flow open? Maybe a watchdog or some other system running in parallel checking for a heartbeat or a max amount of time water can be flowing?
Good point. In my case the program was so simple and the risk low enough that this wasn't needed. The worst thing thar could have bappened was some minor water damage to an exhibition space.
Also my track record of writing stable, bug free embedded software has been pretty solid as of now. But if human life would be on the line (for example) special precautions like multiple independent failsafe mechanisms are non-negotiable.
There are also irrigation controllers that use a ball valve and will work fine under gravity pressure e.g. raised rain barrel. Powered via battery that lasts months and months. No home assistant of course but water usage can be estimated from expected flow and programming.
I hate the tipping culture in the USA and Germany. Instead of being an extra, it feels like an obligatory surcharge I have to pay just to receive the service, good or bad. I usually don’t return to restaurants or bars that nag for tips. In those few places that I like and visit regularly, I don’t give tips. Me being a regular customer brings them more revenue than any tip I’d give otherwise.
Somehow, employers of these establishments convinced the staff that it's the customer’s fault that their wages are inadequate and that they should go after the customers to get the difference. I would much rather pay a higher price and not hear anything about the tips.
I've been playing this one for a long time now. You can play both on mobile and on the web: https://nonograms-katana.com/ the game has quite big community.
Depends if your career depends on some facts not being true. Scientists can seem like a threat to you specifically if for example you need Climate Change to not be real. The last thing you would want is someone bringing evidence and analysis to that reality.
It is good for EU but I belive he was pointing to these hurr durr emigrants bad people. Usually the same people which conveniently always forget that they probalby come as much poorer people than these ones.
I don't see how this is an AI-specific issue or an issue at all. We solved it already. It's called software development best practices.
> A diff can show what changed in the artifact, but it cannot explain which requirement demanded the change, which constraint shaped it, or which tradeoff caused one structure to be chosen over another.
That's not true... diffs would be traceable to commits and PRs, which in turn are traceable to the tickets. And then there would be tests. With all that, it would be trivial to understand the whys.
You need both the business requirements and the code. One can't replace the other. If you attempt to describe technical requirements precisely, you'll inevitably end up writing the code, at very least, a pseudocode.
As for regenerating the deleted code out of business requirements alone, that won't work cleanly most of the time. Because there are technical constraints and technical debt.
And then '90s DSDM (under various stripped-down flavors clustered around agile claiming to be True Agile™) turned into basically WGLL spanning 2 decades going into LLMs:
Note that DSDM purports to "fix" cost but not through estimation per se, but rather by flexing the backlog cutoff:
“DSDM fixes cost, quality and time at the outset and uses the MoSCoW prioritisation of scope into musts, shoulds, coulds and will not haves to adjust the project deliverable to meet the stated time constraint.”
Cost is just headcount, quality should be in your + user's success criteria, and time is (generally) driven by some real-world requirement (event, opportunity, runway, competition, whatever). Varying scope means you didn't plan and roadmap every task up front.
Most everything since are variations on this, tailoring to the needs of the variant's author.
Doing all of this in text-as-code (Markdown, Mermaid, etc.) makes it machinable. Any number of shops were already doing this in text-as-code before the LLMs, giving them a spec-driven LLM context leg up.
To me, the biggest appeal of the Framework laptop is that I can repair it myself and buy OEM parts directly.
I currently own a Lenovo Legion laptop. Still, a very powerful machine, but the screen now has a spot in the middle with multiple dead pixels, the topcoat on the trackpad is peeling off, and the main body has spots where palms rest. I'd happily buy replacement parts and install them, but I can't.
I don't understand the argument you can buy Lenovo OEM parts pretty easily? Even if something is not available through the pcparts site I ordered a replacement display via support.
Yeah! I am also surprised. I have a lenovo from 2015 that's gotten it's wifi card, power IC, RAM - all replaced at some point for very cheap across multiple cities in India. And all this is on a Ideapad. One of their budget "professional" laptops, not even a Thinkpad.
While I understand what Framework is doing and the repairability aspect, somehow this conversation always seems to make it seem Laptops are similar to Ipads or something. It's not.
I wonder how much variation there is between a person who does certain mental activity regularly vs a person who rarely does it.
If they were to measure a person who performs mental arithmetic on a daily basis, I'd expect his brain activity and oxygen consumption to be lower than those of a person who never does it. How much difference would that make?
It involved going to the lab and practicing the thing (a puzzle / maze) I would be shown during the actual MRI. I think I went in to “practice” a couple times before showing up and doing it in the machine.
IIRC the purpose of practicing was exactly that, to avoid me trying ti learn something during the scan (since that wasn’t the intention of the study).
In other words, I think you can control for that variable.
(Side note: I absolutely fell asleep during half the scan. Oops! I felt bad, but I guess that’s a risk when you recruit sleep deprived college kids!)
I worked in an fMRI lab briefly as a grad student. I suspect you'd be correct but perhaps not exactly why you'd expect. Studies using fMRI measure a blood-oxygenation-level-dependent (BOLD) signal in the brain. This is thought to be an indirect measure of neural activity because a local increase in neural firing rate produces a local increase in the need for, and delivery of, oxygenated blood.
The question then is, do you expect a person who is really good at mental arithmetic to have less neural firing on arithmetic tasks (e.g., what is 147 x 38) than the average joe. I would hypothesize yes overall to solve each question; however, I'd also hypothesize the momentary max intensity of the expert to peak higher. Think of a bodybuilder vs. a SWE bench-pressing 100 lbs for 50 reps. The bodybuilder has way more muscle to devote to a single rep, and will likely finish the set in 20 seconds, while the SWE is going to take like 30 minutes ;)
.svg){kind=link}
