firasd

13 ms·

1. 211.▲
   by firasd 1y ago

   Hi folks--I put this together because I'm always frustrated by the 'digital entropy' of quotes online. It's surprisingly hard to find a 100% faithful transcription of a specific movie scene or a passage from a book. This
2. 212.▲
   'What Does a Scanner See?' – Keanu monologue in a Scanner Darkly vs. PKD Book(firasd.substack.com)

   4 points by firasd 1y ago|1 comments
3. 213.▲
   by firasd 1y ago

   Thanks. Basically I find it useful to talk to ChatGPT or Claude after pasting this in.. by now, in comparison, Google Maps feels cumbersome to use for the Delhi metro because of all the tappity-tap-tapping required to pull up routes etc Als
4. 214.▲
   Delhi Metro Text Map, Context-Engineered for LLMs(github.com)

   2 points by firasd 1y ago|2 comments
5. 215.▲
   Delhi Metro Text Map (For AI/ChatGPT Assistance)(github.com)

   1 points by firasd 1y ago|0 comments
6. 216.▲
   The Ghost of ChatGPT 4o(firasd.substack.com)

   2 points by firasd 1y ago|0 comments
7. 217.▲
   by firasd 1y ago

   Interesting! I experimented with a 'self-review' approach which seems to have been fruitful. E.g.: I said Lelu from The Fifth Element has long hair. GPT 4o in chat mode agreed. The GPT 4o in self-review mode disagreed (reviewer wa
8. 218.▲
   I still use GPT 4o(firasd.substack.com)

   1 points by firasd 1y ago|0 comments
9. 219.▲
   by firasd 1y ago

   I wrote this to argue that the GPT-5 rollout backlash wasn't just users hating change, but a rational response to a functional product downgrade. My thesis is that "personality is utility." The collaborative "vibe"
10. 220.▲
    The Ghost of ChatGPT 4o: I told the retired AI model 'people missed you'(firasd.substack.com)

    1 points by firasd 1y ago|1 comments
11. 221.▲
    Qoder Quest Mode: Task Delegation to AI Agents(qoder.com)

    55 points by firasd 1y ago|4 comments
12. 222.▲
    by firasd 1y ago

    Recently I started delegating small bug fixes and feature requests to a coding agent (Qoder) running locally on my repo. In my setup, it drafts specs collaboratively, edits files, runs unit tests, builds the project, and verifies results be
13. 223.▲
    by firasd 1y ago

    Sure. But you'd want to help out the LLM with a message count like this is message 40, this is message 41... so when it hits message 50 it's like ahh time for a new summary and call the memory_table function (cause it's execu
14. 224.▲
    by firasd 1y ago

    A big problem with the chat apps (ChatGPT; Claude.ai) is the weird context window hijinks. Especially ChatGPT does wild stuff.. sudden truncation; summarization; reinjecting 'ghost snippets' etc I was thinking this should be up to
15. 225.▲
    by firasd 1y ago

    > (Disclosure: The Globe and Mail style guide mandates the use of en dashes, which is why you won’t see em dashes used here.) What does this mean? You can't use an en dash as a 'dash'. It's for specialized application
16. 226.▲
    by firasd 1y ago

    I've been doing some really fun things with MCP First of all I think this kind of localhost/stdio MCP is kind of not 'The Way' besides for playing around. I've been working on SSE/remote/cloud-based MCP He
17. 227.▲
    by firasd 1y ago

    Well isn't Wikipedia one of the biggest SPARQL-queryable DBs out there? They use 'slug' based IDs I don't oppose randomization but what I say is that Time is entropy. Use it as a prefix. And UUIDs v7 which uses time is s
18. 228.▲
    Against the UUID(github.com)

    8 points by firasd 1y ago|16 comments
19. 229.▲
    by firasd 1y ago

    The OOD thing ironically helps a bit because sometimes AI can skim over regular timestamps and say two different times are the same etc. LLMs don't actually parse and calculate the timestamps after all. With AlphaDec it's very cle
20. 230.▲
    by firasd 1y ago

    Hi folks, my project: AlphaDec is a human-readable, lexically sortable UTC encoding. It's designed to be timezone-agnostic, reversible, and safe for use in LLMs, filenames, and database keys. Examples like 2025_M4O2_todo.txt carry timi
21. 231.▲
    AlphaDec: A human-readable alternative to ULID/Snowflake IDs(github.com)

    47 points by firasd 1y ago|18 comments
22. 232.▲
    by firasd 1y ago

    Hope you guys find this interesting. Btw here's an interesting demo of AlphaDec's AI-friendly nature.. Claude reverse-engineering it! 1. https://firasd.github.io/alphadec/assets/2025_O3U4_alphadec_... 2.
23. 233.▲
    AlphaDec: A readable, lexically sortable time format for humans and AI(github.com)

    1 points by firasd 1y ago|1 comments
24. 234.▲
    Show HN: AlphaDec, a readable, lexically sortable time format for humans and AI(github.com)

    1 points by firasd 1y ago|0 comments
25. 235.▲
    ChatGPT output 'loop' when prompted to translate the conversation(twitter.com)

    1 points by firasd 2y ago|0 comments
26. 236.▲
    Clean Together: vision-based custom GPT to reduce mental load in tidying up(twitter.com)

    2 points by firasd 3y ago|0 comments
27. 237.▲
    by firasd 3y ago

    This is interesting to me because (if I understand correctly) Twitter uses 'universal ids' like tweets, user profiles, messages etc all use the same type of ID so they need big numbers Also I wonder what underlying factors would m
28. 238.▲
    Twitter glitch shows 18 quintillion bookmarks(twitter.com)

    4 points by firasd 3y ago|2 comments
29. 239.▲
    by firasd 4y ago

    I'm not sure fine tuning helps with this kinda thing. Because the fine tuning mechanism involves sending the API a bunch of data like {"prompt": "<prompt text>", "completion": "<ideal generat
30. 240.▲
    by firasd 4y ago

    I agree that there's some strangeness to it. Like we are not talking to an entity called 'ChatGPT', basically GPT is an omniscient observer and it's guessing what the conversation between the user and the kind of chatbot