paradite

20 ms·

1. 271.▲
   by paradite 3y ago

   I'd say LLM agent, or multi-agents are still in the very early research/prototype stage. You can tell because there are papers from Microsoft on this but no product: https://www.microsoft.com/en-us/research&#x
2. 272.▲
   by paradite 3y ago

   For anyone who didn't bother looking deeper, the SWEbench benchmark contains only Python code projects, so it is not representative of all the programing languages and frameworks. I'm working on a more general SWE task eval framew
3. 273.▲
   by paradite 3y ago

   This is formalized as having a press release at the start of PRD in product management.
4. 274.▲
   by paradite 3y ago

   Agents still very new and nothing that works for production yet. Specifically on using AI for coding, I wrote about different levels of AI coding from L1 to L5, we are still at L2/L3 stage for mature and production ready tech. Agents a
5. 275.▲
   by paradite 3y ago

   Okay thanks for pointing that out. I figure if they do this, they have to throttle or nert it somehow since it is cheaper than ChatGPT Plus which also gives access to GPT-4.
6. 276.▲
   by paradite 3y ago

   Yeah I've been thinking about this lately. LLMs come and go. Prompt engineering techniques come and go. But eval / labelled dataset is always useful once you built it.
7. 277.▲
   by paradite 3y ago

   I believe https://aider.chat/ is working on RAG for codebase.
8. 278.▲
   by paradite 3y ago

   Yeah for context within the single file, GitHub Copilot is good, I use it all the time. But if you use it to do something across multiple files (DB schema, service, controller, HTML, JavaScript), then it becomes less accurate (precision vs
9. 279.▲
   by paradite 3y ago

   GitHub Copilot uses OpenAI Codex, which is a much older model fined-tuned on GPT-3. Definitely not GPT-4, otherwise it would not be less than $10 a month for constant usage.
10. 280.▲
    AI Coding Evolution and Landscape: L1 to L5(prompt.16x.engineer)

    3 points by paradite 3y ago|0 comments
11. 281.▲
    by paradite 3y ago

    I had the same problem of copying and pasting code into LLM web UI so I built a small tool to streamline the process and add source code into the prompt: https://prompt.16x.engineer/ You can't rely on IDE auto context
12. 282.▲
    by paradite 3y ago

    For me, I use a combination of things and I keep regular backups of everything. Photos: Google Photo Notes: Nextcloud +Joplin Documents and invoices: Google Drive I didn't move photos and documents to Nextcloud because the features for
13. 283.▲
    by paradite 3y ago

    Yes. I can't believe the amount of negative comments on Devin or just brushing it off. It is hard to build something like that. If it's just a GPT wrapper, we would have tons of indie hackers building competitors and making 10k MR
14. 284.▲
    by paradite 3y ago

    Sadly I use Macbook with Android.
15. 285.▲
    by paradite 3y ago

    For something that you can download and try right now, and actually works for daily coding tasks, you can try my desktop app 16x Prompt. https://prompt.16x.engineer/ It's not 100% automated but saves a lot of time spen
16. 286.▲
    by paradite 3y ago

    FYI, the correct method is created_at: timestamp('created_at').defaultNow(), // Add created_at column definition Which Claude 3 Sonnet correctly produces. ChatGPT Classic (GPT-4) gives: created_at: timestamp(&
17. 287.▲
    by paradite 3y ago

    I see the purple icon with "ChatGPT Classic" on my share link, but if I open it in incognito without login, it shows as green "ChatGPT". You can try opening in incognito your own chat share link. I use ChatGPT Classic, w
18. 288.▲
    by paradite 3y ago

    I just tried one prompt for a simple coding task involving DB and frontend, and Claude 3 Sonnet (the free and less powerful model) gave a better response than ChatGPT Classic (GPT-4). It used the correct method of a lesser-known SQL ORM lib
19. 289.▲
    Tell HN: Use ChatGPT Classic Instead of ChatGPT

    3 points by paradite 3y ago|0 comments
20. 290.▲
    by paradite 3y ago

    Nice. I build AI-driven puzzle games where you can pick algorithms, tune parameters, or train ML models to play the game. Chess is on my priority list for the next game. Now I know that ML strategy is on par with search algorithms.
21. 291.▲
    ChatGPT for Coding Tasks: Best Practices(16x.engineer)

    1 points by paradite 3y ago|0 comments
22. 292.▲
    Ask HN: Is Cloudflare dashboard having an outage?

    4 points by paradite 3y ago|2 comments
23. 293.▲
    Show HN: 16x Prompt – Coding Prompt Composer(prompt.16x.engineer)

    1 points by paradite 3y ago|0 comments
24. 294.▲
    by paradite 3y ago

    It's just a case of sampling bias. You don't see startups that are not like this because they receive less publicity and press.
25. 295.▲
    by paradite 3y ago

    The info on 24 to 48 hours waiting time is wrong. I just submitted a few new pages (not yet indexed). They got indexed in less than 4 hours.
26. 296.▲
    by paradite 3y ago

    Currently using Joplin with self-hosted NextCloud WebDAV. It has a couple of minor UI/UX issues, even with the plugins. Anyone tried both? How does Obsidian compare against Joplin?
27. 297.▲
    by paradite 3y ago

    Huh? Literally I've seen people on HN making the exact opposite arguments about companies bending over rules for certain markets only to lose the market eventually, hence considered a bad example.
28. 298.▲
    by paradite 3y ago

    I'm sure someone will make a website for it. Sooner or later.
29. 299.▲
    by paradite 3y ago

    I think the budget is used figuratively here to mean where your contribution / value-add to the company is.
30. 300.▲
    by paradite 3y ago

    Since we are taking about license which should be accurate and clear in terms of definitions, "AI" might not be the right word to use. Broadly speaking, AI encompasses much more than machine learning or deep learning. Wikipedia gi