HNHacker News
TopNewBestAskShowJobs

361994752

159 karma · joined July 13, 2017

submissionscomments
361994752··on Show HN: HackerBooks – HackerTube, but for books mentioned on HN
I remember there were multiple such websites before being shared in HN. I bookmarked them, and a couple years later when I finally wanted to find a book, the only thing I got is 404. I hope this one lives longer.
361994752··on You could have invented transformers from memory
Who? Me?
361994752··on Using AI to reverse engineer hardware
I call this wetware mcp
361994752··on OpenCode 2.0
The background task / subagent feature and the ability to specify a model when starting a subagent are so far my favorite improvements. Small things but make huge QoL difference in everyday development.
361994752··on GPT-6 Astra Helped Me Jailbreak My Kindle
I have the opposite experience.

Whenever I start checking a dump, the conversation with Sol won't last five rounds before it flags the work as cybersecurity-related. No matter what I do, the conversation gets cut off in the middle.

Astra is even worse. It starts saying nonsense and even refuses to run a test that we know will cause an AV (as a repro for a nullptr case).

Whenever that happens, I just switch to Opus to continue my work.

361994752··on Why are coding agents so dumb?
I'm using opencode v2, if it matters. I don't have any special skills / prompts for subtasks. Instead, I found new models (I usually have sol or opus as my main model) have the tendency to trigger subtasks, even to a level I need to ask it "please do this task XYZ yourself". The only special prompt I gave is just something like "please use gpt sol or luna for subtasks, base on your judgement". And in my experience it can handle more than 5 async subtasks out of the box
361994752··on Why are coding agents so dumb?
I double checked the publish date before writing down this comment. Because I use the same harness (Opencode, to be specific) as the author, and some of the features are just right there. Like opencode can start multiple subagents for different tasks in parallel. Also I usually ask the main model (e.g. opus 5.5) to pick subagent models for me, and it has no problem identify the difficulty of work and delegate large portion of them to gpt-luna.
361994752··on Triple-A Minesweeper
the non-stopping narration makes it so hard to actually sweep the mines....
361994752··on Man discovers his parents' coffee machine used 1TB of data in 10 days
So you won't want the coffee machine over smart you at that time...
361994752··on GPT‑6 and Intelligent UI for everyone
I really hope they don't merge chat and work. That's kinda the only edge they have over Anthropic at this time point...
361994752··on Show HN: Open-source model routing for coding agents at Astra-level performance
Ofc :) Asking this because we have copilot at work (assuming it is quite common for enterprise)
361994752··on Show HN: Open-source model routing for coding agents at Astra-level performance
Sounds very interesting. Do you have plan to support gh copilot as model provider?
361994752··on ASML currently sells no chipmaking machines in Europe, executive says
then just hire them in asia.. and stick to globalization...?
361994752··on Show HN: Jev Plays Pokémon Red
watched it stuck at rocket hideout for 10 mins.... let me check 1hr later to see if it can find a way out
361994752··on Claude discovers a novel enzyme system with CRISPR-like repeats
and they can explain it in whichever why they like
361994752··on Hister: A private search engine for the pages you visit and the files you keep
There should be no real vendor lock in in my opinion. You can ask the same question with pi + qwen (or any harness + good enough model) with internet access and it will work. Chatgpt is just one option came in handy.
361994752··on Hister: A private search engine for the pages you visit and the files you keep
No I don't share anything with chatgpt. But I do have a $20 subscription, if it matters. I feel It's just capable enough to find what I want from my usually vague and inaccurate description.
361994752··on Hister: A private search engine for the pages you visit and the files you keep
I had the same problem for a very long time but it is largely solved now. I started to simply ask chatgpt "hey I read something about x, y month ago but can't find it now". There is a surprisingly high chance chatbot can just give the exact answer back to me, usually with extra interesting reading materials as a plus.
361994752··on Training a 4B model to produce 81% faster query plans than Postgres
Good and bad plans can have orders of magnitude performance difference. The bad one can easily do enough damage cutting the performance in half before it is canceled.
361994752··on How much oil-market buffer is left?
Since oil companies will make good money from the price hike and exports, I guess they’ll lobby against an export ban?
361994752··on DeepSeek v4.1 flash runs 23 seconds/token on a 2020 16gb M1 Mac Mini
I fell into the trap (again)
361994752··on FDA clears blood test to aid evaluation for Alzheimer's disease
I lost my grandma because of this. She was taken care of by my uncle and my father, and deep in my heart, part of me blames them for not doing their best. At the same time, I know they did way more than I ever could have done. It’s just so hard to take care of someone like that.
361994752··on The road to ACID transactions in Cassandra 6
Why is java bad here...? Because of GC?
361994752··on Understanding is the new bottleneck
Yes. That is what trust means. And being able to trust the right folks makes a good manager.
361994752··on Understanding is the new bottleneck
I’ve worked on both sides, so I know they’re actually very different.

As a manager, the first thing you do is get to know your people. Some of them will be very strong and trustworthy. You give them the hardest work, and you ask them the least. That’s how you scale your team’s scope without getting overburdened. And that’s why teams have key people.

But when you work with LLMs, you still need to understand most of the ideas yourself if it’s a serious product. Because in the end, it means nothing to “trust” an agent. You’re the one responsible for what you and the LLM ship.

361994752··on Gemini 3.7 Flash
I guess the demand is just too high... But even after the price hike, ds is still much cheaper?
361994752··on Mea Culpa – Dark Hours
Did you submit the other half? Which definitely would prove "dog ate my homework"
361994752··on AirLLM 70B inference with single 4GB GPU
I thought I met some huge breakthrough on computer science until I saw your comment
361994752··on 2x, not 10x: coding with LLMs in 2026
Now you actually have very long meetings with AI, if you consider all these long design / explorer / post-poc conversations "meetings".
361994752··on Kimi K3-256k
so Deutschland is not affected. Germany is
Page 1 of 2Next →