alexhans

10 ms·

91.
by alexhans 8mo ago
I just shared this in HN https://news.ycombinator.com/item?id=47026263 to see if it's possible to scale the knowledge sharing and simple and good practices which keep people in control. It may or may not address the pr
92.
Show HN: AI-Evals.io – Evaluate this site with the tools it reviews(ai-evals.io)
5 points by alexhans 8mo ago|0 comments
93.
by alexhans 8mo ago
Congrats on the journey and desire to learn. There's different books or approaches to learn more depending on your needs. It's usually useful to tailor things to whatever you already produce for others or consume for yourself beca
94.
by alexhans 8mo ago
To address the issues of an automated entity functioning as a detractor? I don't think I can answer that specifically. I can brainstorm on the some of the dimensions the book talks about: - societal norm/moral pressure shouldn&#x
95.
by alexhans 8mo ago
When I saw Spark my mind went to Apache Spark and wondered if we were learning all the lessons in orchestration of driver/worker and data shuffling from that space.
96.
by alexhans 8mo ago
This is such a powerful piece and moment because it shows an example of what most of us knew could happen at some point and we can start talking about how to really tackle things. Reminds me a lot of liars and outliars [1] and how society c
97.
by alexhans 8mo ago
High level, the approach is: - I'm pain point driven: - I can't compare what I can't measure. - I can't trust to run this "AI" tool to run on its own - That's automation, which is about intentio
98.
by alexhans 8mo ago
Yeah, the idea is clear. You're "integrating early" and "failing fast" and once you've understood enough about the problem you can design and optimize the right custom tool to make it more accurate, consistent
99.
by alexhans 8mo ago
I'd argue that the moment one cares about accuracy and blast radius, one would natural want to reduce error compounding from a combination of LLM calls (non deterministic) and it's very natural to defer to well tested determinist
100.
by alexhans 8mo ago
Isn't the best eval the one you build yourself, for your own use cases and value production? I encourage people to try. You can even timebox it and come up with some simple things that might look initially insufficient but that discom
101.
by alexhans 8mo ago
> Has anyone tried this yet? If not, why not? Yes (in a minimalistic TDD like way?). [1] I agree that the company as code can be tested without too much risk if you think about as "model your world drawing a circle around what you o
102.
by alexhans 8mo ago
You're right. Control is the big one and both privacy and cost are only possible because you have control. It's a similar benefit to the one of Linux distros or open source software. The rest of your points are why I mentioned A
103.
by alexhans 8mo ago
What are your needs/constraints (hardware constraints definitely a big one)? The one I mentioned called continue.dev [1] is easy to try out and see if it meets your needs. Hitting local models with it should be very easy (it calls API
104.
by alexhans 8mo ago
Useful tip. From a strategic standpoint of privacy, cost and control, I immediately went for local models, because that allowed to baseline tradeoffs and it also made it easier to understand where vendor lock-in could happen, or not get too
105.
by alexhans 9mo ago
There was a post not long ago about a HN user who wanted to both advocate and help people out of this danger: https://news.ycombinator.com/item?id=46662304
106.
by alexhans 9mo ago
Hah! Thanks for sharing. Great to see them doing fun stuff. Just watched a quick yatzhee one. I did play a few in the Ben Jordan series and duty and beyond.
107.
by alexhans 9mo ago
Agreed. Even in the case of a single agent, the compounding of errors [1] can easily make your "flow" unacceptable for your use case. The deterministic where possibe/decoupled/well tested approach is key. With such a fa
108.
by alexhans 9mo ago
Among other things, this got me dabbling with software development. It had an awesome C++ like scripting language. It also used to have a button with a text like "make a game" where you would click what you wanted in the form of c
109.
by alexhans 9mo ago
Agreed. For me a good help, a dry run and a readme with good examples has been the norm for work tools for a while. It's even more relevant now that you can get the LLMs/CLI agents to use your deterministic CLI tools.
110.
by alexhans 9mo ago
> Curious what peeple are doing in the real world It's hard to answer this as the right thing to do might depend a lot on the environment. My pitch would be that automation is not a new problem and the fundamentals still apply: - We
111.
by alexhans 9mo ago
I haven't had the analysis paralysis problem because I've always been quite decent at restructuring environments to avoid bureocracy (which can one of the most dangerous things for a project) but one thing I've observed is th
112.
by alexhans 9mo ago
The fact that you have to dig through things to find it should give you good hints in what is desired for your own code. You want your WHY comments very contextual and findable. 1. Either literally as a comment in the code. Immediately fin
113.
by alexhans 9mo ago
A quick reaction is that there's almost 2 different intents that need to be considered here: - We want to build a business opportunity around auditing. - We want to reduce the amount of insecure Sugabase apps. They align somewhat but d
114.
by alexhans 1y ago
Unless I'm misremembering, aider with Playwright [1] works with local models, so you can scrape the web. Depending on your hardware you could do something like: aider --model "ollama_chat/deepseek-r1:14b" --editor-model
115.
by alexhans 1y ago
Sure. Just sent you an email.
116.
by alexhans 1y ago
Amazon – Economic Decision Science | SDE II | Full Time | ONSITE London, UK We’re hiring a Software Development Engineer for our team at Amazon: Economic Decision Science (London, UK). We're working in an extremely interesting space a