Sometimes I feel that I am not suited to group discussions. Everyone puts forward their own one-sided views, which are unoriginal and cannot stand up to close scrutiny.

Fram2’s Mission Commander @satofishi is set to fly aboard Starship’s first interplanetary human spaceflight mission → spacex.com/updates#first-sta…

The web3/crypto space taught me a harsh lesson as someone who adheres to a long-term investment philosophy. Project teams might simply choose to abandon the project; they won't stick with you and stick to long-term principles. Furthermore, there's little responsibility for failed projects in the crypto space; the project team can simply move on to the next one. This is what I've learned over the past few years.

My personal AI programming experience😂: 1. Cursor programming is like manual driving. 2. Claude Code local programming is like semi-autonomous driving. 3. Codex remote control automated tasks are like autonomous driving.

I have tried using ios app to control mac to code. It would be better if we can also use codex app in one mac to control another mac running codex app.

I have been using hermes and codex app to program while sleeping. Let’s see which workflow is better.

I feel more tired with ai coding: 1. More things want to do 2. Do much more brain context switch than before 3. No deep think like before, just arrange things

One small observation, for Claude Code, debugging a bitcoin node using a remote server seems smoother than debugging locally because it doesn't require frequent permission requests.

LISP 1958, LISP 1.5 1960, PDP-1 LISP 1964, PDP-6 LISP 1966, MACLISP 1969, Lisp Machine Lisp 1975. BBN Lisp 1966, InterLisp 1973. Common Lisp 1984, CLOS 1988, ANSI CL 1994. Scheme 1975 (Revised 1978, 1985, 1986, 1988/1991, 1998, 2007(ouch), 2013...). T 1982.

I am pleased to announce ✨mine✨, a complete, no-frills IDE for Coalton and Common Lisp. It's a brand new IDE that has the essential interactive programming workflow all Lispers talk about, built-in. It is the easiest way to get started with Coalton/CL for Win/Mac/Linux. 👇

I am currently experimenting with using Opus/GPT to assist in designing and optimizing protocols (to see if it is feasible). Compared to using these tools for writing code, the level of uncertainty is significantly higher—primarily because it is impossible to empirically verify whether the protocol functions correctly—and consequently, the volume of manual review required is much greater.

A recent realization: Claude Code is my true teacher.

I've met many smart people that are slow-smart or Bohr-smart. They often don't have a quick answer, they struggle to contribute in meetings (esp. with loudmouths present), but give them a day or week, and they come back with the most brilliant, insightful thoughts and answers.

Can AI coding automatically optimize code quality? For example, can it optimize repetitive and verbose code without human interacting with it? I've recently been manually giving Claude feedback to improve his code, as it was originally too verbose. However, I'm unsure to what extent I need to review his code—perhaps all of it? Or is reviewing it unimportant, as long as it just "works"? Perhaps it's also related to the programming language I use? I use the less common Coalton(common lisp), which Claude isn't yet proficient with. Perhaps Rust would be much better?

Just removed the OpenSpec workflow in my hobby bitcoin-lisp project. The main reason is that I want to use Bitcoin source code for specs, rather than ambiguous markdown files.

Claude code often requires writing Python scripts to check if the brackets in his Common Lisp programs match. This is much less efficient than using Emacs manually. Are there any optimization methods?

I recently tried writing a Coalton-based Common Lisp Ethereum SSZ library using Claude and Codex (initially with almost the same prompts). Then I compared them to my previous "hand-written" Common Lisp implementation. Claude and Codex both used different approaches to their type systems and polymorphism implementations. The image shows a rough comparison of their differences. I personally prefer Claude's approach. Interestingly, I asked Claude and Codex about their preferences, and they both liked the other's version. Gemini, however, voted for Claude.