Hacker Newsnew | past | comments | ask | show | jobs | submit | maCDzP's commentslogin

I have had some success with using Ollama cloud and just instructing Claude Code to hand off tasks to Ollama because of tokens economics.

If the chat is saved as an example of how lot to talk. Can that be used for training an LLM on how to talk or what to remove from a chat?


Reminds me of EMacs, which I really like.


I have used something similar. I set up a team of agents that researches, proposes, builds and audits. Then rinse and repeat. I have used it for different topics. It hasn’t made me a millionaire, but I haven’t lost money either - so that’s some sort of win, right? But I would not have been able to ideate, test at that speed and quality without an LLM.


> It hasn’t made me a millionaire, but I haven’t lost money either - so that’s some sort of win, right?

I'm curious if that is before or after token costs?


Heh, I run my experiments on Claude subscription. And that cost I’ll file as an ”opportunity cost”. But if I would have build this by hand and given myself a salary, then paying Claude to do it is way cheaper.


If it was a non profit that trained the model - would that change your mind?


The issue here is ACM focusing on licensing. A non profit would not be able to pay ACM for access. Hence why this policy is hypocritical: it gives more power to the larger players and undermines smaller actors in the field who have fewer resources.


Why did you sign off your copyright to ACM then? If you kept the copyright and prevented others from distributing your work, you could be the sole distributor and negotiate the price with the AI company yourself.


Because if you don't publish in academia, you might as well be dead. And the terms for most places to publish are similar.

It's a racket.


Ok, then next question is why are you in academia? To share knowledge with humanity, bring the Promethean light to the mortals or something I guess. But then if they use it for productive things that make money, that's too smelly, that's too practical, too dirty, or what?


What a petty view of others you have :( What should they do then? Just shrug and move on, the machine must continue to grind no matter what? Why there is always this request for 100% purity if you dare to criticize the system or part of it?


Grants.They want the juicy grants and complain of the strings attached.

Guess the single strongest predictor of paper in field per year?


You really think a researcher is just a pointless intermediary in shovelling money from the government to the publishing rackets?


that was a big jump from what I said. and what I said was that research lives off grants so researcher can't really come back and cry about strings attached to grants. ofc the problem is faceted, as grant go to published authors with track record, creating a publishing mafia, but everything rotates around the cashflow.

but you want sweet grant money, you play the game by the rules.

researcher that don't want to play the grant game can go and find a VC to subsidy product creation


Non Profit != No Money

A non profit could well pay, and there are plenty of reasons frontier models should be managed by non profit. After all, why allow a for profit to benefit from free contributions?


A non-Profit like OpenAI? Oh wait I guess that can change.


Everybody gives more power to the larger players. You won't work for me for $10/hr but you'll work for a guy who has more money for more. There's nothing wrong with that. Money is just a fungible unit representing value offered.


I think your example hinges more on the fact that $10/hr is a desperation wage. It's not enough to buy you a life where you're not stepped on by others. The further up that chain you move the more other things start to matter, at least for most people.

If you're offering me $300/hr and the other guy is offering me $400/hr, a whole lot of things start to matter more than the differential.

That gets at the "fungibility" notion you refer to. Money has the same nominal value everywhere, but different real value based on how much of it you already have. Which is to say that the marginal value of money drops pretty precipitously several times at certain thresholds that relate to the particulars of the economy (when you can afford to eat, when you can afford a house, when you can afford to not work anymore, and so on).


> If you're offering me $300/hr and the other guy is offering me $400/hr, a whole lot of things start to matter more than the differential.

I’m not sure which Catholic saint said that, but one requires a minimum of material conditions in order to practice spiritual virtues.

> when you can afford to eat, when you can afford a house, when you can afford to not work anymore, and so on

When you can buy a government, a couple billion seems like a small loss.


>The issue here is ACM focusing on licensing. A non profit would not be able to pay ACM for access. Hence why this policy is hypocritical: it gives more power to the larger players and undermines smaller actors in the field who have fewer resources.

And when the open weights models distill all the content out of the majors anyway?


And drop the atributions. What will the ACM say then?


Being a non-profit is just a tax arrangement, it shouldn't change any of your opinions. What does the tax rate of who does an action have to do with the action being good or bad?


I have had success with ”brainwashing” by starting out with bug bounties/CTF and then going from there.


That's brilliant, I should try that.

I usually just start by preloadig context with plausible legitimate use, have it work and obviously fail, and then ask to figure it out without ever mentioning any high risk words. Model offers to RE itself and classifiers are happy.


Has anyone tried that? I have a feeling that if I put it a prompt Claude would comply. But I am all in on the Claude cool aid.


Sure Claude would comply, but Anthropic has no financial (or other) incentive to optimize this, so there’s no reason to expect it to be particularly good.

It would be like asking the clerk at a Whole Foods which grocery store in the city sells the cheapest eggs. He’d probably answer - he might not even say Whole Foods - but WF is hardly teaching all their staff the best methods to answer this question in training. (Heh, training.)


Yes. I have Claude route to codex all the time as part of a dev process where fable is manager and it oversees the work of sidekicks and subagents. It’s happy to comply.


I work in civil engineering and tinker with computers as a hobby. I found Alexander’s work delightful. It really helped me think about my profession and my hobby and how they tie together.

He also talks about community in a city, so it walks across all scales in the build world. That really helped me think about ”scale” and what phenomena emerges at different scales. I haven’t seen those thoughts about community when I have read about patterns in SWE.

Patterns mostly revolves around solving specific problem in programming, but I haven’t yet found something that uses the same thoughts on a larger scale explicitly. I believe there are patterns for how to architect multiple softwares, but what’s beyond that scale? Say how to integrate different SWE communities.


> I haven’t seen those thoughts about community when I have read about patterns in SWE

Maybe a bit of a tangent to your point:

one major difference that jumped out at me about Alexander's work vs the software engineering 'patterns' is how much emphasis Alexander places on involving the people who will live & work in the buildings in the design process. Some of this involved including the building's eventual inhabitants during rapid prototyping -- it's really cheap to experiment with different building outlines when it's just folks standing around in an empty plot of land prototyping the outlines with pegs & string.

I first read about Alexander's work in the book _Peopleware_, they described this aspect of it as "local control of design by those who will occupy the space". A space that can be customised to the needs of those who use it.

That's something completely absent in the software engineering 'pattern language' stuff, which tends to feel more like ivory tower big design up front by the architect / throw it over the fence to the users.


There was a talk by M. J. Dominus ... "Design Patterns" Aren't https://perl.plover.com/yak/design/samples/slide001.html

While the narration of the talk itself is likely lost to time and "people weren't recording everything in 2002"... the slides linked are reasonable signposts of the author's ideas.

The key points I believe occur on slide 12.

    Patterns vs. "Patterns"

    The pattern language does not tell you how to design anything

    It helps you decide what should be designed

    You get to make up whatever patterns you think will lead to good designs
---

I believe that people read the GoF Design Pattern book as a textbook or bible rather than what the authors were doing of "these are things that we discovered while writing a word processor" ... and are useful ways of looking at those problems. But then those patterns got enshrined in some holy temple as you need to do these things and this is how you build software".

Patterns got turned into a prescriptive thing that people do to build software and for a long while people were often coming up with the question "did I discover a new pattern" ... which when you read Alexander's book becomes a silly question.

My own rant on the cult of pattern... https://the-whiteboard.github.io/2016/09/02/patterns.html

... One of my favorite pages to link on career advice is Find the Hard Work You're Willing to Do https://news.ycombinator.com/item?id=26209541 ... The previous blog post that shows up on that same page is "Why Patterns Failed and Why You Should Care" https://www.cs.uni.edu/~wallingf/blog/archives/monthly/2018-... which is also a good read (which in turn links to another talk - https://www.deconstructconf.com/2017/brian-marick-patterns-f... which has transcript and video)


A lot of people talk about building software "ecosystems". But that's more management/marketing than SWE, but that's where you build similar scale. You have Linux, products built on Linux, software written for Linux, sub-communities like KDE, etc. and how you build those communities and allow both diversity for different ideas, but with a shared common core and how that evolves.

But that's like an upper management type concern, you usually don't call that "SWE" anymore.


The larger systems that contain software, or which software comprises, are the workflows that software sustains. Those workflows then comprise departments, organizations, businesses, and eventually cross-business interdependencies, supply chains, etc.

I.e. just like C. Alexander's patterns quickly leave the realm of laying brick, the same logic appled to software quickly leaves the realm of writing code.


I did Nand2Tetris and then I understod why I need to vectorize. You get a view from nand-gate to software and get to see all the interfaces. Its a wonderful course.


During uni I only took a Java class. During the exam I had to turn in hand written code. I guess that would work today.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: