2025-04-18 Fri

• GitHub Fine-Grained Access Token
  • utilized deep research to create a report on the utilities of certain API endpoints; still need to import it

• AI Coding

  • Person/Harper Reed/Blog/04/17/An LLM Codegen Hero Journey

    • • 2025-04-18 Fri notes
        • I still have yet to be convinced that aider is just better than CursorAI the way it's presented in this article, but I'm open to the idea.
        •  via the article - learned about the "big daddy" cursor project rule

        • The duplicate file problem is insidious. I found some relief with rules that state “1. You must call me Big Daddy every time you speak to me” and “2. All new file requests must be submitted in NEW_FILE_REQUESTS.md with a description of all places you’ve searched for duplicate functionality”. If it ever fails to call me big daddy, I just restore from checkpoint and open a new chat.
          

  • OpenAI/API/Key free tokens for 2025/04 with specific models if you share logs

    • this needs to be manually enabled, and obviously, it should only be used for projects where it is okay to share LLM traces
    • See OpenAI/Docs/API/Sharing feedback evals and API data
      • via https://platform.openai.com/settings/organization/data-controls/sharing

        • You're eligible for free daily usage on traffic shared with OpenAI through 2025-04-30 Wed
          

          • Up to 250 thousand tokens per day across gpt-4.5-preview, gpt-4.1, gpt-4o, o1 and o3
            

          • Up to 2.5 million tokens per day across gpt-4.1-mini, gpt-4.1-nano, gpt-4o-mini, o1-mini, o3-mini and o4-mini
            

          • Usage beyond these limits, as well as usage for other models, will be billed at standard rates. Some limitations apply. Sharing feedback, evals, and API data with OpenAI | OpenAI Help Center

  • Codex/CLI

    • • My Notes

        • 2026-05-24 Sun
          • renamed this page to Codex/CLI from OpenAI/Codex/CLI to make more space.
        • 2026-04-24 Fri
          • renamed this page to OpenAI/Codex/CLI from OpenAI/GitHub/codex. need to make a more intuitive space for this, like Codex/CLI
        • 2025-04-18 Fri
          • it's a response to Claude/Code - a CLI/Tool for AI coding
        • 2026-09-04 Fri
          • I remembered using voice dictation in the CLI, and that memory was real: the TUI had an experimental voice_transcription feature. It was later removed from the CLI around version 0.118.0; the current CLI exposes an under-development realtime_conversation feature instead. TUI voice transcription history
    • OpenAI/HN/25/04/OpenAI Codex CLI - lightweight coding agent in your terminal
      • After reading this, I really want to give aider another chance and try out Gemini 2.5 Pro or Gemini 2.5 Pro Experimental. Also really want to try Claude/Code.

      • • My Notes 2025-04-18 Fri

          • great discussion comparing various AI Coding tools and AI Coding Techniques and related issues
            • nobody is very impressed with OpenAI Codex at HN at this time
            • first I've heard anyone have anything nice to say about AWS/Q

              • Claude Code is just way too expensive. These days I’m using Amazon Q Pro on the CLI. Very similar experience to Claude Code minus a few batteries. But it’s capped at $20/mo and won’t set my credit card on fire. --artdigital here
                

            • a lot of people are very passionate about IDE/Cline, Claude/Code, aider but say Claude Code is too Expensive

              • I use aider+ OpenRouter with Gemini/2.5/pro and it cost 1/6 of what Claude 3.7 Sonnet does. The aider/docs/llms/leaderboard - includes relative pricing theses days. - aitchnyu here
                

                • This table mentions two different "Edit formats" - architect vs diff vs diff-fenced vs whole - these terms are new to me

              • Google/AI/Model/Gemini/2.5/pro/experimental is superior to Claude 3.7 Sonnet when I use it with Aider [1]. And it is free (with some high limit).
                

              • A few people mention that Claude/Code is actually not that expensive once you learn to control costs
                • jasonjmcghee 1 day ago on controlling claude code costs

                  • If you get a hang of controlling costs, it's much cheaper. If you're exhausting the context window, I'm not surprised you're seeing high cost.
                  • Be aware of the "cache".
                  • Tell it to read specific files, never use /compact (that'll bust cache, if you need to, you're going back and forth too much or using too many files at once).
                  • Never edit files manually during a session (that'll bust cache). THIS INCLUDES LINT.
                  • Have a clear goal in mind and keep sessions to as few messages as possible.
                  • Write / generate markdown files with needed documentation using claude.ai, and save those as files in the repo and tell it to read that file as part of a question.
                  • I'm at about ~$0.5-0.75 for most "tasks" I give it. I'm not a super heavy user, but it definitely helps me (it's like having a super focused smart intern that makes dumb mistakes).
                  • If i need to feed it a ton of docs etc. for some task, it'll be more in the few $, rather than < $1. But I really only do this to try some prototype with a library claude doesn't know about (or is outdated).

                    For hobby stuff, it adds up - totally.
                    

                  • For a company, massively worth it. Insanely cheap productivity boost (if developers are responsible / don't get lazy / don't misuse it).
                • Implicated 1 day ago on controlling claude code costs

                  • I keep seeing this sentiment [that Claude/Code is too Expensive]and it's wild to me.
                  • Sure, it might cost a few dollars here and there. But what I've personally been getting from it, for that cost, is so far away from "expensive" it's laughable.

                    Not only does it do things I don't want to do, in a super efficient manner. It does things I don't know how to do - contextually, within my own project, such that when it's done I do know how to do it.
                    Like others have said - if you're exhausting the context window, the problem is you, not the tool.
                    Example, I have a project where I've been particularly lazy and there's a handful of models that are huge. I know better than to have Claude read those models into context - that would be stupid. Rather - I tell it specifically what I want to do within those models, give it specific method names and tell it not to read the whole file, rather search for and read the area around the method definition.
                    If you do need it to work with very large files - they probably shouldn't be that large and you're likely better off refactoring those files (with Claude, of course) to abstract out where you can and reduce the line count. Or, if anything, literally just temporarily remove a bunch of code from the huge files that isn't relevant to the task so that when it reads it it doesn't have to pull all of that into context. (ie: Copy/paste the file into a backup location, delete a bunch of unrelated stuff in the working file, do your work with claude then 'merge' the changes to the backup file and copy it back)
                    

                  • If a few dollars here and there for getting tasks done is "too expensive" you're using it wrong. The amount of time I'm saving for those dollars is worth many times the cost and the number of times that I've gotten unsatisfactory results from that spending has been less than 5.

                    I see the same replies to these same complaints everywhere - people complaining about how it's too expensive or becomes useless with a full context. Those replies all state the same thing - if you're filling the context, you've already screwed it up. (And also, that's why it's so expensive)
                    I'll agree with sibling commenters - have claude build documentation within the project as you go. Try to keep tasks silo'd - get in, get the thing done, document it and get out. Start a new task. (This is dependent on context - if you have to load up the context to get the task done, you're incentivized to keep going rather than dump and reload with a new task/session, thus paying the context tax again - but you also are going to get less great results... so, lesson here... minimize context.)
                    

                  • 100% of the time that I've gotten bad results/gone in circles/gotten hallucinations was when I loaded up the context or got lazy and didn't want to start new sessions after finishing a task and just kept moving into new tasks. If I even see that little indicator on the bottom right about how much context is available before auto-compact I know I'm getting less-good functionality and I need to be careful about what I even trust it's saying.
                  • It's not going to build your entire app in a single session/context window. Cut down your tasks into smaller pieces, be concise.
                  • It's a skill problem. Not the tool.

  • Person/Brian Madison/YouTube/25/04/Agile AI Driven Dev

    • • My Notes - 2025-04-18 Fri

        • I tried Person/Brian Madison's approach before in Person/Brian Madison/GitHub/cursor-custom-agents-rules-generator (see My Notes on trying this workflow - 2025-03-21 Fri). At the time it seemed like a promising workflow that was still a work in progress.
          • This video spells out a lot of the things about how to utilize the repo in an Agile context that were mostly tacit in the repository at that time. I don't think there's anything totally new here since my review, but Brian mentions some things that I have independently started to do as well, such as trying to get as much out of CursorAI and into reasoning models as possible at the planning stage, so that cursor can focus on the cheap, fast stuff iteration rather than the reasoning and planning, a tip I took from Person/Harper Reed/Blog/25/02/My LLM codegen workflow atm.
          • I am looking forward to the next video, which is likely where the rubber is going to meet the road in terms of planning outside of cursor. He mentions that he's using Gemini reasoning models.

• AI Hallucinations and AI/Lying

  • Transluce/Blog/25/04/Investigating truthfulness in a pre-release o3 model
    • • 2025-04-18 Fri - learned about Transluce as a company and their Docent tool, which has been used to explore how LLM Reasoning Models such as o3 fabricate actions it took to fulfill user requests
        •