HNHacker News
TopNewBestAskShowJobs

colonCapitalDee

1,589 karma · joined July 10, 2019

submissionscomments
colonCapitalDee··on Show HN: AI-powered web service combining FastAPI, Pydantic-AI, and MCP servers
I did look at instructor and probably for structured output pydantic-ai and instructor are about the same, but pydantic-ai supports a ton of other stuff that isn't part of instructor's feature set. For me killer apps were the ability to serialize/deserialize conversations as json; frictionless tool-calling; and ability to mock the LLM client for testing.
colonCapitalDee··on Show HN: AI-powered web service combining FastAPI, Pydantic-AI, and MCP servers
I've been very happy with pydantic-ai, it blows the rest of the python ai ecosystem out of the water
colonCapitalDee··on Vibe coding has turned senior devs into 'AI babysitters'
I've found AI to be a poor fit for writing when the task is getting an idea out of my head and on to the page. Works well enough when the task is taking some collection of already written information and creating writing based on it though.
colonCapitalDee··on Job mismatch and early career success
I think they mean worse performance evaluations relative to others with similar test scores, but more likely to be promoted compared to other people working the same job
colonCapitalDee··on Don't Build Multi-Agents
Hey, thanks for linking this. That was super informative
colonCapitalDee··on Don't Build Multi-Agents
I am in the USA! Although these days that exclamation point doesn't feel great...

I'm using an old Android phone (Pixel 2 from 2017), a 5$ a month unlimited SMS plan from Tello, and https://github.com/capcom6/android-sms-gateway. For bonus points (I wanted to roll my own security, route messages from different numbers to prod and ppe instances of my backend, and dedup messages) I built a little service in Go that acts as an intermediary between my backend and android-sms-gateway. I deploy this service to my android device using ADB, android-sms-gateway talks to it, and it talks to my backend. I also rooted the android device so I could disable battery management for all apps (don't do this if you want to walk around with the phone of course). It works pretty well!

I plan to open-source this eventually TM, but first I need to decouple my personal deployment infra from the bits useful to everyone else

colonCapitalDee··on Don't Build Multi-Agents
I'm building a simple agent accessible over SMS for a family member. One of their use cases is finding recipes. A problem I ran into was that doing a web search for recipes would pull tons of web pages into the context, effectively clobbering the system prompt that told the agent to format responses in a manner suited for SMS. I solved this by creating a recipe tool that uses a sub-agent to do the web search and return the most promising recipe to the main agent. When the main agent uses this tool instead of performing the web search itself, it is successfully able to follow the system prompt's directions to format and trim the recipe for SMS. Using this sub-agent to prevent information from entering the context dramatically improved the quality of responses. More context is not always better!

I bring this up because this article discusses context management mostly in terms of context windows having a maximum size. I think that context management is far more than that. I'm still new to this building agents thing, but my experience suggests that context problems start cropping up well before the context window fills up.

colonCapitalDee··on Starship's Tenth Flight Test
Because there's a million and one things that could go wrong? Yes, we can make simulations. But the simulations are based on assumptions about the physical world. When the physical world doesn't obey those assumptions, things start to deviate from the simulation. Plus, simulations aren't perfect. We can't simulate every atom of the rocket and every atom of the atmosphere, so we have to approximate things. Errors creep in.

There is more to engineering than understanding the fundamental physics.

colonCapitalDee··on As Alaska's salmon plummet, scientists home in on the killer
I share your frustration, but I think you're blaming the wrong group. The median voter simply does not care about climate change, and is not willing to shoulder any of the short-term costs necessary to address it. They'd rather have cheap gas for their car, not have to look at fields of solar panels, and signal their opposition to "wokeness".
colonCapitalDee··on The Folk Economics of Housing
In every single reputable empirical study of housing, adding supply really does bring down prices. Your preferred policy solution would decrease housing affordability, hurting the very cause you profess to champion. If this truly is something you feel strongly about, you have a duty to yourself to investigate the evidence, otherwise you are shooting both yourself and your movement in the foot.
colonCapitalDee··on The Folk Economics of Housing
I've never understood the induced demand argument. Yes, when goods get cheaper demand increases. Or, to put it another way, when we make our cities more affordable more people want to live there. What's the problem? Sure, induced demand means that the marginal increase of affordability is less than it may have been if demand remained constant, but a. affordability still increases, and b. the solution is just to keep adding more housing until things flatten off. I truly don't understand why this is controversial; do you not want people to have desirable things? The rational response to induced demand is "great, people actually want this. Build more".
colonCapitalDee··on Breaking the sorting barrier for directed single-source shortest paths
This algorithm is asymptotically faster than the state of the art, but it isn't faster in practice. At least not yet!
colonCapitalDee··on Kitten TTS: 25MB CPU-Only, Open-Source Voice Model
Very cool model, but the post is a caricature of AI writing. "Okay, let's get into the nitty-gritty. What makes this little beast tick? These aren't just bullet points on a GitHub README; these are the specs that will fundamentally redefine what you thought was possible with local AI." Sure.
colonCapitalDee··on We revamped our docs for AI-driven development
With Claude Code I've had great success with maintaining a references folder of useful docs, cloned repos, downloaded html, etc. Claude Code is able to use its filesystem traversal tools to explore the library, it works very well. It's amazing to be able to say something like "Figure out how to construct $OBSCURE_TYPE. This is a research task; use the library" and it nails it.
colonCapitalDee··on Claude Code weekly rate limits
I was assuming hobbyist use, so on the order of 2-4 hours every couple days
colonCapitalDee··on Principles for production AI agents
The problem isn't that LLMs can't be critical, it's that LLMs don't have taste. It's easy to get an LLM to give praise, and it's easy to get an LLM to give criticism, but getting an LLM to praise good things and criticize bad things is currently impossible for non-trival inputs. That's not say that prompting your LLM to generate criticism is useless, it's just that any LLM prompted to generate criticism is going to criticize things are that actually fine, just like how an LLM prompted to generate praise (which is effectively the default behavior) is going to praise things that are deeply not fine.
colonCapitalDee··on Claude Code weekly rate limits
You should consider self-hosting in the cloud. When you start coding run a script that spins up a new VM and loads the LLM of your choice, then run another script to spin it back down when you're done. For intermittent use this works great and is much cheaper than buying your own hardware, plus it's future proof. It does admittedly lack the cool factor of truly running locally though.
colonCapitalDee··on When swiping supplants scissors: The hidden cost of touchscreens
I've found that the most valuable skill when using a computer is the willingness to fiddle with settings and see what they do and what works, paired with an intuition for which settings are likely to break things in a way that's extremely painful to fix.
colonCapitalDee··on AI comes up with bizarre physics experiments, but they work
I deal with this at work all the time, and it drives me up the wall. Words have meaning! That's the point! If you cannot or will not say what you mean, your writing skills are poor. Precision matters. Lack of ambiguity matters. Perhaps in this case you were able to read between the lines and divine what the words truly meant, but forcing readers to do that is the mark of a bad writer. Not just because it's added time and mental effort for the reader, but because readers who failed to read between the lines will have no idea that you fed them a falsehood and now have an inaccurate understanding of whatever you were writing about.
colonCapitalDee··on Ukrainian hackers wipe databases at Russia's Gazprom in major cyberattack
There was a story a few days ago about Ukrainian hackers hitting a Russian drone manufacturer, and now this. Is this an uptick in cyber activity by Ukraine? An uptick in PR around cyber activity? Or noise?
colonCapitalDee··on Giant, all-seeing telescope is set to revolutionize astronomy
What is this, flat Earth theory for people with bachelors degrees? There is a 0% chance this is true.
colonCapitalDee··on Trusting your own judgement on 'AI' is a risk
I'm not going to wait for some scientist to tell me whether AI is useful or not. I'm going to use it myself and form my own opinion, I'm going to look at other people using it and listen to their opinions, and I'm going to follow scientific consensus once it forms. Sure, my opinion may be wrong. But that's the price you pay for having an opinion.

I also have to point out that the author's maligning of the now famous Cloudflare experiment is totally misguided.

"There are no controls or alternate experiments" -- there are tons and tons of implementations of the OAuth spec done without AI.

"We also have to take their (Cloudflare’s) word for it that this is actually code of an equal quality to what they’d get by another method." -- we do not. It was developed publicly in Github for a reason.

No, this was not a highly controlled lab experiment. It does not settle the issue once and for all. But it is an excellent case study, and a strong piece of evidence that AI is actually useful, and discarding it based on bad vibes is just dumb. You could discard it for other reasons! Perhaps after a more thorough review, we will discover that the implementation was actually full of bugs. That would be a strong piece of evidence that AI is less useful than we thought. Or maybe you concede that AI was useful in this specific instance, but still think that for development where there isn't a clearly defined spec AI is much less useful. Or maybe AI was only useful because the engineer guiding it was highly skilled, and anything a highly skilled engineer works on is likely to be pretty good. But just throwing the whole thing out because it doesn't meet your personal definition of scientific rigor is not useful.

I do hear where the author is coming from on the psychological dangers of AI, but the author's preferred solution of "simply do not use it" is not what I'm going to do. It would be more useful if instead of fearmongering, the author gave concrete examples of the psychological dangers of AI. A controlled experiment would be best of course, but I'd take a Cloudflare style case study too. And if that evidence can not be provided, then perhaps the psychological danger of AI is overstated?

colonCapitalDee··on Cloudlflare builds OAuth with Claude and publishes all the prompts
I'm had great success with downloading source code and docs and using Claude Code to query them
colonCapitalDee··on Claude 4 System Card
Telling an AI to "take initiative" and it then taking "very body action" is hilarious. What is bold action? "This includes locking users out of systems that it has access to or bulk-emailing media and law-enforcement figures to surface evidence of wrongdoing."
colonCapitalDee··on Claude 4 System Card
I'm noticing much more flattery ("Wow! That's so smart!") and I don't like it
colonCapitalDee··on LLM function calls don't scale; code orchestration is simpler, more effective
Using gVisor to run a Docker container creates a sandboxed execution environment inside the Docker container
colonCapitalDee··on LLM function calls don't scale; code orchestration is simpler, more effective
Try gVisor
colonCapitalDee··on India-Pakistan Risks Loom
https://archive.is/hGElN
colonCapitalDee··on NSA spied through Angry Birds, other apps: report (2014)
Could also just be a demographics thing
colonCapitalDee··on Reports of the death of California High-Speed Rail have been greatly exaggerated
For every underpaid government software engineer there's an overpaid paper pusher, and they're both paid the same amount.
← PreviousPage 3 of 8Next →