Indie developers report big time savings using Claude Code

I calculated the Pentagon's "black budget".... Using Claude, the LLM banned from the Pentagon by Hegseth, I went through the military’s initial budget submissions. I found $126 billion in programs either explicitly marked as classified, or described only through a code word. (1/2)

Burned $91.34 with Claude Code /goal in 3.5 hours Unreal, It was able to reverse engineer my Amaran bluetooth lights so I don't have to use the crappy app

“People are becoming what we now refer to as AI Vampires…an individual programmer uses Codex or Claude Code or one of these AI coding systems…They’re working just like more hours than ever, and then the AI Vampire thing is literally this thing where they stop sleeping.” - Marc Andreessen

This site's #1 anthropic glazer cannot do it anymore :( this really really does fucking suck a lot. I think later I'm going to take a crack at building some sort of wrapper around non-`-p`'d Claude code to get it running non-interactive. This fucking blows!

#Quarto is much more than a publishing framework. When leveraged with agents with Claude code, it's a powerhouse tool for agentic workflows. In this blog post, I make a few quick arguments for why you should check out Quarto if you're working with agents. www.jhelvy.com/blog/2026-05...

Strongly believe that the proliferation of Claude Code Skills in real world codebases is gonna get Bitter Lesson’d soon

The rise of ai in software development actually makes REAL software engineering more critical than ever. While tools like GitHub Copilot, Claude Code, and Cursor make ai code generation cheap, they don’t solve the real bottlenecks. Video now on @modernswe.bsky.social ➡️ youtu.be/XDNXLdwq114

Right now you can run Claude code in a non-interactive mode for cool things like automation (like GitHub actions), bots or other stuff that doesn’t go through the normal chat flow, and that usage applies to your monthly usage. The new change will track that usage differently

Claude subscription is now not able to be used for any programmatic use of Claude Code, including use in GitHub Actions, use of -p, or any 3rd party Claude Code GUI. This is after OpenAI has embraced allowing Codex subs to be used anywhere. Tokens always cost the same. Anthropic just wants lock in.

Use one of the thousand link in bio sites or…just have Claude code one for you. Took 2 prompts

this is why i feel ok using claude code to develop a food allergy tracker that doesn’t exist and and i’d never have energy to do otherwise it’s gonna help people and after thorough code reviews (it’s only 1500 lines of sinatra so far), i’ll oss it but that help is worth $100/mo to me, not $2k

“I have gotten stuck on bug fixes where, when I run out of Anthropic tokens in Claude Code, I couldn't work anymore. The current system I am working on started to become a monstrosity of complexity where I didn't even know what most of it does anymore…”

Compass vulnerability research identified code execution paths affecting AI coding assistants including Claude Code, OpenAI Codex and Cursor. The findings will be demonstrated live at @thezdi.bsky.social Initiative #Pwn2Own Berlin 2026, May 14 to 16. #AIsecurity #LLM

claude code certainly does a good bit of cache optimization but it'll still eat through that usage fast in anything other than a pretty occasional workflow.

One of the reasons genAI is generally crappy is because it's all regex and Python, all the way down. We've seen how crappy Anthropic's code is; it's believable that Claude wrote more of it as they claim, because it sucks. genAI is inefficient in every measure. But ML doesn't have the "push button,