npiano
npiano on Hacker News
Why is that simple or plausible? Why is simpler or more plausible than lifting an almost-finished solution from a researcher's account?
This is a sci-fi trope with no practical or realistic grounding. To "run", the AI needs vast banks of interconnected GPUs. These are pretty easy to spot and don't fit into anyone's pocket.
It very clearly didn't do what they said they told it to do. We do not know what was truly intended.
Because there is no internal feedback loop. There is no internal process gathering information from the external world and using it to modify its future behaviour. There is only the beginning of the calculation and then the end.
Posted by
npiano on Monday June 08, 2026 @03:30AM
from the show-hn dept.
https://www.ai-boost.io/
Hi HN,My number one frustration with LLM agents is that every session starts blank and I have to re-explain my patterns and conventions every time.I often find myself doing things like "look at X repo and copy the patterns used for Y", which requires the repo to be on the local machine.A couple big ones for me personally:
- Auth flows
- Terraform patterns for AWSRules and skills help to some...
EMPs only occur in very high altitude tests
A genuine question, presuming no correct answer: what is to be done about it? China is reportedly on track to run more than 50% of global manufacturing by 2030, if the World Bank is correct. What would you do to act against this? Is doing nothing acceptable?
This assumes that logic is derived from predicting the next step from previous information, which is not accurate.
Yeah so many people (clients, and I guess contractors) don't realise this. There's a lot of work to keep a site going.
Posted by
npiano on Sunday January 12, 2025 @05:50PM
from the ask-hn dept.
I only recently started doing contract work, and the first few times have been pretty smooth, but it occurs to me that handing someone a repo with working code in it with instructions is a bit naïve. What solutions do you use to prevent full use without full payment?
It's true it's not unique. I would be interested to know what you believe are the main reasons why it fails. Thanks!
Thanks! Honestly, I think this approach is inevitable given the rising tide of unstoppable AI spam.
I would be interested in people's thoughts here on my solution:
https://www.tela.app.
The answer to bot spam: payments, per message.
I will soon be releasing a public forum system based on this model. You have to pay to submit posts.
Location: Valencia, Spain
Remote: yes
Willing to relocate: no
Technologies: React, Django, FastAPI, Ethereum, Bitcoin, AWS
Résumé: https://www.linkedin.com/in/nicholas-piano/
Email: nicholas [dot] piano [at] protonmail [dot] com
Currently wrapping up construction phase of Tela Network. Long experience in web design and implementation along with deployment and system design. Have experience with payment integration. Open to contract opportunities in related areas.
I always think of the holographic orbit mapping UI from The Expanse when I see something like this. It's the paper of the future that will be hooked into everything we think about. Such a powerful tool for exploring the world.