Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
whazor
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
6 ms
·
121.
▲
by
whazor
6mo ago
The FAQ says: > We might modify your laptop to remove or power down the battery But powering down battery is not enough against the fire risk. Servers get hot 24/7 and might still overheat the battery.
122.
▲
by
whazor
6mo ago
Imagine you are the top engineer of your company. Everybody wants your attention, many meetings, design sessions, and of-course code reviews. With Claude Code, I use Gitlab for reviewing code. And then I let Claude pull the comments. It loo
123.
▲
by
whazor
6mo ago
Obviously hardware wise the real blocker is memory cost. But there is no reason why future devices couldn't bundle 256GB of mem by default.
124.
▲
by
whazor
7mo ago
There is also https://microvm-nix.github.io/microvm.nix/ if you want increased isolation.
125.
▲
by
whazor
7mo ago
I have found that maximising AI coding is a skill on its own. There is a lot of context switching. There is making sure agents are running in loops. Keeping the quality high is also important, as they often take shortcuts. And finally you n
126.
▲
by
whazor
7mo ago
There are also software model checkers that can model distributed processes. You have to simplify the state a bit, otherwise you get a state space explosion. I tried it out myself, I let AI add action transitions through the code, like: &#x
127.
▲
by
whazor
7mo ago
I wonder if something like Gemini Veo (AI video generator) could re-make old movies. It might be hard to keep the consistency over the entire movie.
128.
▲
by
whazor
7mo ago
I found that it normally takes one prompt early-on to go from 'vibe-coded spaghetti' to something having a decent architecture.
129.
▲
by
whazor
7mo ago
Chatbots already have memory, and mine already knows my schedule and location. It doesn't even need to say anything directly, maybe just shorter replies, less enthusiasm for opening new topics. Letting conversation wind down naturally.
130.
▲
by
whazor
7mo ago
One of the most reliable ways to induce psychosis is prolonged sleep deprivation. And chatbots never tell you to go to bed.
131.
▲
by
whazor
7mo ago
There are (illegal) ways to delay trains as a person
132.
▲
by
whazor
7mo ago
Don’t forget the requirement of not dropping frames under load. The browser engine might have assumed that requirement throughout the entire code base.
133.
▲
by
whazor
7mo ago
With agentic loops, you specify what you want and it continues to do stuff until ‘it works’. Then publish. Its takes less time and attention. So projects are less thought out and less tested as well. In the end, I think it’s not about how a
134.
▲
by
whazor
7mo ago
This direction could be an interesting AI benchmark. All kinds of different humans use LLMs for their job, whether allowed or not. Including diplomats, defence personnel, lawyers etc etc. Within the benchmark you could play both sides and r
135.
▲
by
whazor
7mo ago
This is about increasing output per machine via upgrades.
136.
▲
by
whazor
8mo ago
This is how I solve the dependencies with Nix: https://gist.github.com/whazor/bca8b687e26081e77d818bc26033c... Nix helps Claude a lot with dependencies, it can add stuff and execute the flake as well. I will come back
137.
▲
by
whazor
8mo ago
I push my branches daily, so I wouldn't lose that much work. If it breaks then I ask it to fix it. But I do quickly check the output what it does, and especially the commands it runs. Sometimes it throws all code in a single file, so I
138.
▲
by
whazor
8mo ago
I currently have a TUI addiction. Each time I want something to be easier, I open claude-code and ask for a TUI. Now I have a git worktree manager where I can add/rebase/delete. As TUI library I use Textual which claude handles qu
139.
▲
by
whazor
8mo ago
what if you add a <!-- see /llms.txt --> to every .html
140.
▲
by
whazor
8mo ago
OpenClaw is more like an art project than a consumer product. It has shown clear consumer product demand. The next step is making it a safer consumer product.
141.
▲
by
whazor
8mo ago
A big challenge will be the unfamiliarity of such a new system. Many people have found that coding agents work really well with terminals, unix tooling, and file systems. It's proven tech. Where-as doing DB queries to navigate code wou
142.
▲
by
whazor
8mo ago
is there any such manager out there that uses the ACP protocol? ( https://agentclientprotocol.com/ ) Ideally I would like a ACP proxy wrapper, where I integrate ACP into my code editor and still be able to use it remotely via
143.
▲
by
whazor
8mo ago
or SAP
144.
▲
by
whazor
8mo ago
Kimi 2.5 is an opensource model with many providers https://openrouter.ai/moonshotai/kimi-k2.5/providers Sure, opus and codex are significantly better. But price wise they cannot deviate too much from open models.
145.
▲
by
whazor
8mo ago
Drinking coffee gives positive health effects. Drinking too much coffee and getting over-caffeinated has bad effects.
146.
▲
by
whazor
8mo ago
I want a work/gaming monitor that looks like a Samsung frame. We have a little desk in the living room and I want to hang a monitor on the wall, that shows pictures when not using it.
147.
▲
by
whazor
8mo ago
The lethal (security) trifecta for AI agents: https://simonwillison.net/2025/Jun/16/the-lethal-trifecta/
148.
▲
by
whazor
8mo ago
oh my the security risks
149.
▲
by
whazor
8mo ago
ASML did two extremely big bets on the future. Both futures are now.
150.
▲
by
whazor
8mo ago
Concretely, when you use Django, underneath you have CPython, then C, then assembly, and finally machine code. I believe LLMs have been much better trained on each layer than going end-to-end.
More ›