HNHacker News
TopNewBestAskShowJobs

sgt101

7,940 karma · joined January 24, 2012

nothing to see here, move along, move along.
submissionscomments
sgt101··on ChatGPT Developer Mode: Full MCP client access
I love how we're getting to the Neuromancer world of literal voodoo gods in the machine.

Legba is Lord of the Matrix. BOW DOWN! YEA OF HR! BOW DOWN!

sgt101··on How can England possibly be running out of water?
It was privatized in the 1980's not 1960's.
sgt101··on How can England possibly be running out of water?
I wondered if anyone would point this out. England's population has risen 25% in my lifetime. Worse, the (perfectly adequate for this level of population) water infrastructure in in the wrong place... It's all in the North East in order to supply steel mills that are no longer there!
sgt101··on The maths you need to start understanding LLMs
could you give an example of "HN would not like this AI leetcode"?
sgt101··on I bought the cheapest EV, a used Nissan Leaf
There's something wrong with your friend's model 3. If I stand on my accelerator it goes straight (if set straight) and the suspension (set on sport) is hard.

It's not M5 BMW hard, but it's hard. I guess that if you have a performance variant then maybe it would be.

sgt101··on I bought the cheapest EV, a used Nissan Leaf
EV's look like EV's because they don't have the same constraints on space that ICEV's do. There is no engine in the front/back, no water pump, no oil pump, no water cooled radiator, no starter motor... so on so on. There's also a great big battery under the floor pan and no gas tank.

They are also heavier, so need different set ups in order to handle decently.

But, that's why they look so different.

sgt101··on Mark Zuckerberg freezes AI hiring amid bubble fears
We should also say that "being really lucky is the best way to make sure that other people don't have as much luck as you do"
sgt101··on Best Practices for Building Agentic AI Systems
Could you produce some evidence that what you are advising is actually useful?
sgt101··on Best Practices for Building Agentic AI Systems
Like the OP article, no data, no evidence - random ideas presented as advice.

I mean, could you even do some anec-data style before and after comparisons? Like, do 5 problems with no explicit request for tests and 5 without and show that you get 0 out of 5 for no test and 5 out of 5 for a test?

People like me might carp on about how that's not good enough or something, but to be honest I would feel quite uneasy and unjustified complaining unless I had done similar work... without this sort of attempt to justify your thoughts though I feel like a school teacher intervening in a discussion where some kid is advocating the use of heroin to her peers on the basis that Aunt Agnes says it feels good.

sgt101··on The AI Job Swindle
In the case of economic failure.

Increased productivity due to automation isn't an economic failure - it's a redistribution challenge.

sgt101··on The AI Job Swindle
>I’m bringing this up because, in the workplace, AI models have the potential to deliver both things. And that should terrify you.

Calm down!

Things that should terrify you:

- bioweapons

- nuclear war

- the collapse of society

- a civilization ending solar flare

- a comet striking earth

- a mega volcano

Gyrations in the job market are worrying, especially for young people, but they should not terrify you.

sgt101··on Kodak says it might have to cease operations [updated]
There's quite a lot of interesting literature on this. Unfortunately the best work looks at Polaroid [1] rather than Kodak so I'm on slightly shaky ground, on the other hand I think that if Clay Christensen can point at Kodak as a exemplar of the same processes [2] it's all ok.

Anyway the issue seems to be that Polaroid (also Kodak) had what's called a razor blade business model. In Polaroids case they actually sold the cameras for the film, but they sold them cheap cheap cheap so as to encourage the use of the high margin product that they really wanted to shift - the film. Kodak went one better (sorta) by becoming a category killer in film so everyone who wanted to use a non-polaroid camera probably ended up buying Kodak film (razor blades).

There isn't really an equivalent consumable in digital, the action is in the camera (which ended up being connected to the smart phone). Now, Polaroid had consumer goods experience and consumer product R&D capability from making instamatics, and they had a management culture of investing big to create the next big thing - but even they just couldn't get past the "I'm telling you there is no business case for this" conversation in the C-suite. And here comes Clay with the Innovators Dilemma. The Polariod and Kodak executives both struggled to see past the investment cases which said things like "invest $100m to get $40m a year from efficiency savings in film manufacture, 100% for sure for ever here is proof that can't be discussed" to see "invest $20m to get $5m next year, growing 200% a year for the next 10 years maybe, I promise, honest that's what the future looks like fellas". Worse than that - investment in the digital industry would clearly kill the film industry, and the future for a digital Kodak looked radically different to the future for a film based Kodak and therefore there was a big constituency that didn't want to see their lovely chemical factories and lab based R&D shuttered in favour of a basement filled with computer geeks, also a bunch of people saw that their million $$ sales commissions for selling to corner store processors would go out of the window.

The turkeys... they do not vote for Christmas.

So the innovation cases that could have saved both companies didn't get the bucks and the rest is history. Sucks to be the shareholder left holding that stock when the veil of corporate propaganda masking the collapse finally fell. I remember a presentation from a Kodak team in about 2009 that would have convinced you that the future was indeed filled with Kodak moments spilling through everyone's life forever. Unless you actually knew what a digital camera was and how good they were then.

In the last few years a new thread of work [3] that unifies these kind of insights into a theory of innovation processes has emerged. I am a big fan but it hasn't got the kind of profile I think it should, probably because of Quantum AI Crypto flaps. It's quite instructive to use it to understand the Quantum AI Crypto stuff as well, but no one likes to hear about that because it's more fun to look at the pretty pictures I guess.

[1] Tripsas, M., & Gavetti, G. 2000. Capabilities, cognition, and inertia: Evidence from digital imaging. Strategic Management Journal, 21: 1147–1161.

[2] Christensen, C., Raynor, M.E. and McDonald, R., 2013. Disruptive innovation (pp. 20151-20111). Brighton, MA, USA: Harvard Business Review.

[3] Grondal, S., Krabbe, A.D. Chang-Zunio, M. THE EVOLUTION OF TECHNOLOGY. Academy of Management Annals 2023, Vol. 17, No. 1, 141–180. https://doi.org/10.5465/annals.2021.0086

sgt101··on Claude Code is all you need
I'm sorry but people who let an agent run on prod deserve what they get. Basically even saying you would do that should disqualify you from working in IT in the way saying "I like to drink when I'm working" should disqualify you from airtraffic control.
sgt101··on How AI conquered the US economy: A visual FAQ
A big dynamic in this is that the business models for two of the biggest players are at stake.

Google is facing a significant danger that its search advertising business is going to just disappear. If people are using AI to find stuff or get recommendations then then aren't using google. Why should a photographer continue to spend $200 a month on ads when their clients are coming from openAI? Especially when OpenAI is using the organic search results. Meta is facing the same sort of issue, if the eyeballs aren't on insta then the ad $$$ go somewhere else.

So if they have money, and can get money, they will invest it to protect their businesses - all of it.

MS and VC's are doing the opposite, they are investing with the idea of taking the attention that Google and Meta currently have, but they are also following the "I'm scared" signal that Google and Meta have put in the market.

sgt101··on Ask HN: What trick of the trade took you too long to learn?
Find out exactly what needs to be done.

So often the assumed goals of for an activity are malformed or the result of misunderstandings. Spending time pinning down what has to be done, why, and what the value of it is enables many other questions to be asked such as "wouldn't it be better to do X instead." The answer to that can often be a simple "no, because Y" but sometimes knowing Y is extremely helpful, and sometimes the answer is "yes, you are right" which is much better.

sgt101··on Is the interstellar object 3I/ATLAS alien technology? [pdf]
It's weaponised language that pseudo academics hurl around at each other to try and denigrate the research outputs of other people. In the distant past it had a meaning which was that research was being published in small parts in order to get more academic kudos from it, but now literally all research is published this way based on the judgement of the submitter about what they can get accepted where.

In this case Loeb seems to have decided to delight in publishing out-there ideas, probably with a bit of a mission to open up debate and widen the range of acceptable topics in the field of astronomy for younger less established researchers. Basically, he's at a point in his career where he simply doesn't care what anyone things of him and his research and so he's spending credit so that if someone younger and more at risk than him comes up with a startling idea they will hopefully be more likely to share it.

I think it's a good thing, obviously a bunch of people really don't.

sgt101··on The natural diamond industry is getting rocked. Thank the lab-grown variety
where do you buy them at that price?
sgt101··on The natural diamond industry is getting rocked. Thank the lab-grown variety
Interestingly it's been possible to make rubies for years, 150 years ago rubies were far more prized than diamonds, but then fell out of fashion, mostly because diamonds were in good supply and there was a big marketing campaign for them. Consequently lab grown rubies are pretty cheap - maybe $50 carat, but here's the thing.. natural rubies are still expensive because they are rare and contain impurities that give them a glow/fire that is prized.

But here's the fun thing. You can destroy the value of a natural ruby by cooking it. If you heat it in the right way the impurities will anneal out and you will be left with a near flawless stone that appears to be artificial! Bye bye value :)

sgt101··on Developing with Kiro: Amazon's New Agentic IDE
It's strange because I used replit in an online container and then had to download the code to carry on editing it - so I don't know how it managed to do that! Anyway, it sounds like they were crazy giving the keys to prod...
sgt101··on Developing with Kiro: Amazon's New Agentic IDE
Super interesting that no one here has mentioned replit.

I found the experience of using Kiro and replit really similar with one important difference: replit mostly worked.

Kiro tore off and wrote tonnes of code and tests. It asked me to peck at approval requests (one thing I liked was the regexp type trust this tool to do this requests) it spent half a day creating an app to do what I asked... and it was incomprehensible bullshit. I couldn't do anything with the project and I have not touched it since.

Replit was a bit more interactive, but pretty autonomous, and it got me to 90% of the solution and then stalled out - wouldn't correct some of the problems I identified to it. About 2 hrs with Cursor sorted that out though.

I did use Cursor to do it "AI assisted" and that took about the same amount of time. The advantage is that I really do know whats going on in the code base, but the Replit + Cursor solution is actually better in the sense that it looks better, and works better because the agent did some bits a bit more nicely than I did with Cursor - so I got those ideas for free.

Anyway :

Hand coding = a walk through the wilderness

Cursor = motor cross scrambler bike up the mountain

Replit = a helicopter ride to somewhere higher up the mountain selected at random that you didn't know about but now you have to get to the peak by yourself buddy, good luck

Kiro = you are blindfolded in a container of some sort and it's moving.

sgt101··on Major quantum computing advance made obsolete by teenager (2018)
Her thesis is now available "Quantum Machine Learning without any Quantum"

https://ewintang.com/assets/tang_thesis.pdf

sgt101··on Shale Drillers Turn on Each Other as Toxic Water Leaks Hit Biggest US Oil Field
The peak and decline of US oil production is going to be one of the geopolitical stories of the next 25 years. If US renewables and nuclear don't fill the gap the need for global naval dominance is going to come knock knock knocking again. It looks very like there are three factors that are going to mean that reasserting dominace isn't just very hard:

- Missile tech has advanced and proliferated.

- US ship building is crippled, and expanding it (especially where needed) is going to be very difficult. Basically the US is going to need a lot of SSN and it won't be able to make them.

- There is a competitor power with sufficient resources to make it a real competition, and a very different conceptualisation of how the world should work.

sgt101··on How Tesla is proving doubters right on why its robotaxi service cannot scale
I guess the gamble was that there would be a certain point where the edge cases disappeared into the noise and it didn't matter that the approximation was/is "wrong" because the behavior of the car would match the requirements of the situation even if it wasn't for the right reasons.

To be fair I remember reading about GPT2 and thinking that LLMs would blow out for similar reasons.

sgt101··on How Tesla is proving doubters right on why its robotaxi service cannot scale
Andrej "train it on more data and the problem will go away" Karpathy
sgt101··on Nobody knows how to build with AI yet
So far:

- when I ask models to do defined that I know how to do and can tell them about that method but can't remember the details off off hand and then I check the answers things work.

- when I attempt to specify things that I don't understand fully the model creates rubbish 7 out of 10 times, and those episodes are irretrievable. About 30% of the time I get a hint of what I should do and can make some progress.

sgt101··on The bewildering phenomenon of declining quality
I bought some wrangler jeans, because I remembered that I had a pair years ago that fit me better than levis, the levis were quite expensive and the wranglers cheap.

They were ok for the first wear - but not great to be honest.

Then I washed them and they were unwearable.

Didn't do anything fancy, just a cool wash, dried them on the line.

They turned to cardboard.

sgt101··on Advertising without signal: The rise of the grifter equilibrium
How can consumers respond to get useful information and enable themselves to make better decisions when exchanging their resources for goods?

- use other information channels like review media (fashion sites)?

- trust secondary sites like brand retailers (IE. John Lewis in the UK?)

????

sgt101··on I Never Cared Much for Swords. Then I Had to Fight with One
Maybe it's different for other people, but I just couldn't move the way I used to, I had less gas, less snap, my reactions were poorer. There's a thing of power coming from your gut/core... that's what lets you really move and use leverage, and that really faded. Competing with guys in their mid twenties was the shocker. I guess that there's a think where experience and training builds advantage but then you go over a cliff where the physical difference just overwhelms that.

Look, we don't see professional athletes in their 40's much if at all. Mike Tyson was the best fighter (ever?) and looked in shocking good shape - but he still lost to that goon.

sgt101··on I Never Cared Much for Swords. Then I Had to Fight with One
A guy called Steve Paul coached me for a few years, the connection is that he was Peirce Brosnans double in Die another Day. He got motor neuron disease just before covid and the poor old sod fell down some stairs when he was on holiday and died.

Irony can sometimes be a bit harsh.

Fencing was such a big part of my life for so long, but when I got to my late 30's the power went out of my body. It was shocking, but just true, I couldn't do what I used to be able to do when I was young and I had to come to terms with it. A lot of people go into coaching, or make their peace and fence as a veteran, but I couldn't do that. It took me a long time to grasp why because I used to coach when I was competitive, so why could I not abide it when I knew I could not compete.

The answer was not attractive. I envy the young. I cannot stand to watch them and know I am not one of them.

If you are young then take up swords, or racquets, or gloves and revel in your sinuous power. Soon it will be gone, and all you will have is memory, until something comes sliding and slipping and takes even that, and you find yourself tumbling into the night.

sgt101··on LLM Inevitabilism
Looking up a project on github, downloading it and using it can give you 10000 lines of perfectly working code for free.

Also, when I use Cursor I have to watch it like a hawk or it deletes random bits of code that are needed or adds in extra code to repair imaginary issues. A good example was that I used it to write a function that inverted the axis on some data that I wanted to present differently, and then added that call into one of the functions generating the data I needed.

Of course, somewhere in the pipeline it added the call into every data generating function. Cue a very confused 20 minutes a week later when I was re-running some experiments.

← PreviousPage 6 of 34Next →