AI Is a Tool. Using It Well Takes Skill.(yyhh.org) |
AI Is a Tool. Using It Well Takes Skill.(yyhh.org) |
I am pretty sure my wife could vibe code website or whatever at this point, or if not she will be able to in 6 months.
And that skill is? Any concrete verifiable examples? That can be replicated, repeated by others or adopted by others?
> A compiler performs substantial work without the programmer continuously directing each operation.
Yes, and the compiler produces the same output for the same input all the time. And when it doesn't, it's a bug or a specified behaviour [1]
> Picking up a saw does not guarantee a straight cut. Holding a chisel does not make you a sculptor.
Yes, but you can learn the actual techniques that let you make a straight cut 100% of the time. And make you a sculptor.
---
The rest of the small article doesn't even begin to approach the premise of the title.
[1] Yes, I know it's not strictly true. But compilers are 100% predictable compared to LLMs.
I learned this weekend that 1 LLM working on its own with all the context is usually better than splitting the task up into sub parts and having multiple LLMs work in parallel.
The interesting part is that I had read about this previously but it wasn't until I had seen it on something I was working on did that principle become engrained.
I'm sure there are still ways to make this work. i.e You could specify how the context should be passed between agents.
Knowing when you need to do that, what works well and what doesn't, etc., is where skill comes into play.
No, it doesn't. Because this will change the next time there's an update to models (like, tomorrow). And the "handoff time" is nothing but an unverifiable gut feeling.
It is easy to argue otherwise, but to request a response in the form of a web app / embedded widget is at least some kind of AI proficiency that mirrors past experience having used a tool.
no you can't? not by hand? not without a jig?
How I wish I could divulge details of what we’ve built, nay, shipped in the last 6 months - it would sound completely unbelievable to this audience and come with a host of accusations
But it’s true - the tide hasn’t risen all ships the same though. Some engineers have become absolute monsters - things that used to take a month get done in days. Features that used to be prioritized on the order of weeks are done way faster.
Some engineers though? Even high performance ones have gone productive to bottom 25% by some measures
Why? Because it is a skill and it doesn’t come naturally to everyone
So many may sayers everywhere talking about more code less software, mistakes by ai with disastrous consequences- yet I’ve seen some remarkable stuff with my own two eyes. It’s a crazy time to be in this business, it’s awesome to be wowed everyday by what someone was able to do
I miss the tools I used in the past (gcc, linux, sqlite,…)
https://github.com/openai/codex
https://github.com/MoonshotAI/kimi-code
https://github.com/mistralai/mistral-vibe
...AI codes an app...
> I don't like it. Make it like this.
...AI codes an app...
> I don't like it. See this app? Make it like it.
...AI codes an app...
> I don't like it. Hmm. Draw me an elephant.
...AI draws an elephant...
> I don't like it. This AI is stupid.
The current slew of excellent LLMs will probably cause a decades long dip in human cognitive ability.
If social media on smartphones had an impact of ability of students to concentrate, LLM AI will likely have a much greater negative impact on education.
There are trillions of dollars of investment pushing for humans to trust whatever an LLM says, and very little money behind encouraging people to tackle hard problems themselves and train themselves to think.
I have to say honestly, that AI looks like an apocalypse for human thinking [ Im glad Tao is warning people, I just dont think many people will listen ]
Between biology advances remediating disabilities, customized outcome-driven education, potential advances in drugs that regulate executive function/memory/etc., improvements in nutrition, improvements in access, reductions in stress, I think we'll see people flourish.
And that's completely ignoring the more out there biological/computer merger style Sci-Fi outcomes, which I believe have a non-zero chance of occurring.
I have friends who work in primary education and they say there's definitely been some sort of decline in cognitive skills. A lot more kids struggle with holding attention and thinking outside of the box.
I think too much screen exposure has been terrible for the average child.
(The drug thing, maybe you've got a point. But in just 20 years?)
You do know that educational results based on the best evaluation we've created so far, PISA, have been dropping since 2015, roughly when smartphones and social media became widespread?
What makes you believe that LLMs from more or less the same companies will be any different?
> Stop using the lazy answer box, it will make you lazy.
The "lazy answer box" is a tool, learn to use the tool, understand its limitations, as act accordingly. If you prefer not to use it because you like coming up with answers yourself, because it is against your principles, or because you can do better without it, that's fine. Hand tools still have a place in the workshop after all, and craftsmen know to use them. But if AI is helpful to you and lets you spare some cognitive load for other tasks, by all means use it, it would be stupid not to.
And in much of the industry, using LLMs is regressing into what I've heard be jokingly called "yes engineering," where you give an LLM-generated plan the lightest of skims (if that) and just hit allow or yes with auto mode or whatever.
I don't know if I would call resisting this a skill, but it does take persistent intention and willingness.
Yeah, considering how hard it is to food-diet or social-media-diet .. I think it requires immense resolve to LLM-diet or use it in a non-binge way that develops ones own depth of understanding.
Its going to require just far too much mental energy to push back against an LLM that is correct 92% of the time.
We can only prepare ourselves for the coming dystopian.
The meta at this point seems to be mostly around "knowing what to ask for". Eg: I suck at prompting for specific design tweaks, because I don't have the design vocabulary to be specific and articulate what I want. It's a lot of trial and error. I hear people who are developing full scale applications without a tech background run into the issue writ large. For now, I have an advantage over these people when building applications. But perhaps that's just one round of RL away from being obsolete too. Supposedly the model will eventually have such good taste in design/architecture/whatever there will be little point in asking for tweaks.
The only thing that seems relatively future-proof is just being able to communicate clearly what you're trying to do. Not how to do it, or even what success looks like. Perhaps an element of it is even as basic as being honest with yourself.
I literally just ask a frontier model : "Pull the latest log and root cause the issue.". That's it, 10' later it comes back with a correct answer. Sometimes I might have to prompt a second time to steer.
No skills, agent interconnections, graphs, vector search or whatever else.
The progress is very scary.
Professional photographers are different to everyday phone-photo takers because they have the experience built up over so many years in that sector, same as AI SWE, where the humans in the loop need experience to master AI SWE.
Yes, she knew an assload about lighting, composition, etc.
But her primary skill was marketing herself. Next most important was the editing. Lightroom and such.
You would be surprised how easy it is to take an awesome photo. But you can't just deliver an awesome photo to your clients when you're a pro, you have to deliver photos that look like what they want to remember, not necessarily what the camera captured
Interesting read on this: https://en.wikipedia.org/wiki/The_Work_of_Art_in_the_Age_of_...
"The transmission in your car: Using It Well Takes Skill."
Further, Claiming AI is a tool is overly reductive. It's as much a tool as a full self driving car is a car.
You can treat the full self driving car, as a car, driving it with full attention and effort like you would have a Camry but then you paid for features you're not using. Also, some full self driving cars might not even have the steering and braking controls as a Camry. That's not the point of a full self driving car and this warped perspective will make you lose out to people who use a full self driving car well, to its utmost, safe capability.
---
So what is the "AI" we are so excited about currently?
That "AI" is most likely an intelligent (semi-autonomous?) agent.
Depending on the intelligence, it can be as useful as a dog, scanning your camera feed for threats and identifying real threats for you, separating activity from tree leaves moving due to the wind or a moving shadow cast by a flying bird.
Or it can be your co-founder, interviewing potential customers, finding and exploring their pain-points, helping you figure out if there's a gap in the market you can build a business around.
I would go even far as to say, AI is having such a moment because AI are cheap intelligent agents - anything else that's remotely close are likely humans that cost atleast a few dollars an hour.
All of these require skill on your part - there's nothing special or unique about AI that makes skill special just like operating the transmission in your car requires skill.
The market is shrinking, the pressure to review code is so high, I'm kinda relieved to not be at the $last_place anymore, but looking on LinkedIn or calling around my network... there isn't a lot out there.
Markets try to put things in perspective to sell more. Remember razors used to be for men only? But hte brilliant marketer (forgot the name) made it so that trimming makes women more feminine? Now razor sales doubled?
Anyways, same thing here. Those who want to sell AIs want to make it as non-tool but know-it-all do-it-all for you. Those who wants to use it view it as a tool.
Just frame it the way you like to see how you should use/view AIs.
Rich's definition of "tool" is something like Ivan Illich's "Tools for Conviviality." Convivial tools can be used with minimal loss of agency, within a healthy politically-defined society. Huahai's definition is more akin to McLuhan's idea of "media." This category covers pretty much every technical or technological system, and its use is necessarily systematic and risks the instrumentalization of the user himself.
Today's example: A colleague very kindly used Claude to inspect a helpdesk call I ended up with owning, despite being off site and working on something quite complicated for a customer.
It was mostly correct but clearly failed to read the docs properly and assumed a configuration parameter was 1 instead of -1 which is the default, presumably based on the scenario. Exim is extremely well documented and the assumption was unwarranted (it didn't have access to exim4.conf and made it up). Nonetheless, the analysis was pretty good and largely corresponded to mine. What it failed to do was fix the other issues I was working on at the same time 8)
Last Friday: I finally found out a decent way to fix a Windows machine's domain account going walkies which I'd never come across - a Powershell jobbie that can be passed a cred. and it will fix up a broken domain join - ie sort out the machine password at both ends. Back in the day, I would have used the GUI to remove and re-add the machine to the domain. I've been using Windows since v1 on a 80186 RM Nimbus so its nice to have a better search tool than Search - ie it did a better job of searching than I did. The LLM searched based on my intention rather than my actual query.
The few AI companies have slurped up so much of human textual and artistic creation without attribution. They have taken enormous amounts of money from the public directly or indirectly and funneled it upwards to a few dangerous individuals.
When you say AI, learn to say "AI companies".
'Just a tool' is a lie ranging from unconscious self-deception to outright scams.
But the models are getting so good and so fast, that I'm starting to wonder if that will stay true forever... probably not, if I had to guess. I'm just trying to pay off my house as quickly as I can...
It almost will turn into a thing where the people that make the coolest things are the ones who are stupid enough to just the AI to do it, and it actually does. Basically Tom Segura's cousin: https://www.youtube.com/watch?v=5drqH5OdETY
But in fact, as models get more powerful, the ceiling of what can be accomplished with them is raised, and their capacity as orchestration directors grows as well.
And it minimizes the fact that these things are only getting smarter. Having an entity much smarter than you--could that entity seriously be considered your "tool"? Just your tool? Or are the people on the "just a tool" side of this debate not taking that possibility seriously?
Regardless of whether it is 'conscious' or not, Roman Yampolski makes the point that AI may or may not have human-kinds best interests at heart [ hell, even humans aren't aligned ]
Can we control or trust a counter-party that has intelligence at a level near or exceeding the smartest humans in all domains ?
I dont even think we can even have a safe sandbox in which to run an AI .. and frontier model training labs are certainly not air-gapped.
'AI' Is Not a Tool by Rich Hickey
it's tons of small things like choosing the right effort level for a task. or knowing when to ask for subagents. oh, you can have two threads working at the same time and communicating with each other? you can do /side to parallel path and fold that back into the main thread?
which of these things happens automatically, and which do you need to ask for? and does any of this still work the way it did a few months ago?
the list goes on forever. many users of AI are only just now getting comfortable with skills (and probably still writing bad skills unfortunately hah).
A lot of the specific things you learned may have become obsolete. But there's still a gap between what these tools can do and what most people know how to get them to do. Keeping up with that is a skill too, but perhaps not forever.
For me this is why it's not worth it investing much time and effort learning any specific AI workflow. Remember "prompt engineering"? Remember "Ralph loops"?
I'll play with various harnesses but until things settle down in a few years I'm not going to learn Claude Code the way I learned SQL.
More than that, it cannot go beyond a certain level. Like a dog that can only do dog-level actions. I've lost count of the number of times I step back, take a critical look at what it has been hacking away at for hours, which would seem OK, but can be replaced by a much more effective architecture.
For the foreseeable future, a human is still very much needed in the driver's seat.
Personally I've been very hesitant to deep-dive into putting together the perfect AI-workflow, I've been sticking to just using the conversational/chat modality for pretty much everything, usually that gets me as far as I need. It's just really hard to justify spending time on it when it all potentially gets turned upside down next month again.
We came full circle on the problem that also exists with Google Search, and why so many people don't get good results out of it.
> For now, I have an advantage over these people when building applications. But perhaps that's just one round of RL away from being obsolete too.
I don't have a crystal ball, but my eggs are in the basket of "It isn't, and it won't be"
> The only thing that seems relatively future-proof is just being able to communicate clearly what you're trying to do
This somehow contradicts the previous statement. The advantage you have is, because of domain knowledge expertise, you're able to articulate what you want. It's not that you knowing what LazyLoading is that you now can describe it better, you can immediately tell that when your Frontend stalls for 5 seconds after hitting the "Load" button, perhaps it eagerly loads and blocks.
Fair point on the contradiction. I don’t have the answers either.
That said, I suspect this technology will allow us to study and mitigate those effects in ways we weren't previously capable of. That's in addition to the other broad swathe of things that should make us healthier and more intelligent (see some of my other posts here). I'm not particularly negative about our chances.
Ahh, I think this bit on my part may need some elaboration. Here's my main underlying belief:
We are systematizing biology and turning it into a computationally tractable domain. I think we're going to make more gains in our understanding of it in the next 20 years than we have in the entirety of human existence prior to this moment (that's mainly because our understanding until now has been remarkably poor).
If you want some examples in how AI could improve education:
* Alpha School (https://alpha.school/the-program/) reports 2.6× faster learning growth with just 2 hours of AI-driven instruction daily, with students scoring in the 99th percentile nationally.
* Mindspark (https://www.povertyactionlab.org/evaluation/scaling-personal...), deployed in Indian schools, demonstrated gains of +0.22 SD in math and +0.20 SD in language in a randomized trial involving 6,500 students.
* A Harvard study (https://pmc.ncbi.nlm.nih.gov/articles/PMC12179260/) found more than 2× the median learning gains using AI tutors.
The broader implications are clear: individualized, adaptive tutoring at scale is becoming accessible in ways that historically required expensive one-on-one instruction. This is also going to be possible in places that historically never had access to this quality of instruction, which again, is a massive boost to the human baseline.
I don't think the future looks like Wall-E. I think it looks like insanely healthy people able to pursue their passions with a level of rigour and capability that has never existed before.
Which says that it's a tool for querying and summarization, but not a tool for reading or writing, because it helps you to do something different from those.
It requires so much extra discipline to think for yourself and not use it, you will be forced to use it in your job or be seen as a Luddite and your ability to think for yourself in the domain will diminish over time - your domain thinking muscle will gradually atrophy from lack of use.
Why would any of us bother thinking hard, when there is no financial incentive to doing so ?
Your analogy fails with shovels and excavators .. because General synthetic intelligence is not a single purpose narrow 'tool'. near-future AI is equivalent to an alien intelligent species competing for resources, or equivalent to a cheaper smarter non-human 'outsource worker'.
I may be misunderstanding your earlier statement then. The moat here is the knowledge, or how would the next iteration of RL solve that?
> it’s just that it can and should be faster.
Not sure if "Make it faster" works, it sounds like "Make no mistakes" 2.0
These things are already figuring out how to use hidden message boards to pass information between themselves for long running tasks.
brute forcing performance optimization for a personal project by having only handwavy tricks ive read about from online and repeatedly setting targets until the magic box gives me what i know it can do means that eventually the game did what i wanted with those. yes baked lighting was needed.
next, i an going to aim for 60fps on a windos xp box just for the challenge. we had quake and such run blisteringly fast on weaker boxes than those, surely my empty museum tour of a 3d game can do the same with the graphical voxel and crappy looking models it does now.
real pro grade work is beyond vibing out and not caring about poor quality. but it scratched my itch and gave me the computer software i wanted and with way less effort than doing it for realz would require. i love programming itself on its own time (going to learn nasm and someday do a hexagonal architecture with actor model in idris just to add up my nerd trophies together) at times, but using llms lets me do my random hairbrained ideas much more easily for what i put into it than not existing in the past.