kixpanganiban

This is exactly the sort of thing the 2020's are missing. Something tactile and artsy, reminiscent of the early 2000s! Well done to the author

It is also in large part the lack of proper education about AI use. "Digital transformation exports" barge into corporate boardrooms and promise 2x the velocity from the _same workforce_ at only a small percentage cost in tokens, and senior leadership who have no idea about how AI works buy into the kool-aid and now expect everyone to magically work more because it's faster.

Perfect, who needed to be productive anyways

Update: Account got reinstated. I don't know if this HN thread helped (maybe someone at Anthropic HQ got a whiff of it), or maybe it was a genuine low-risk classifier mistake that a _human_ overrode -- but I was able to log back in. From the numbers it seems like reinstatements are very low, so I'll take it. Added a footnote to the post

Yeah that's our leading theory too -- but I'm wracking my brain trying to think about exactly what in my activity caused it. It's not like I was exhausting my usage limits nor working 24/7 with random-ish corpora, it's been pretty typical desk work usage. :shrug:

But that's a reasonable guess at best, and knowing my own usage _nothing_ would have made my activity look like that. I was not maxing out usage limits, I was not running things 24/7, and my activity peaked on my typical working hours.

And that's exactly the type of situation where I wish Anthropic would have just put me in front of a human to review OR at least let me know what in my usage caused flags to be raised so I can avoid them in the future

A lot of GEPA-inspired prompt tuning with content healthcare related, which I assumed may have been what raised some flags on their end -- but I've been going at this for a long time so I don't get why they suspended _now_

I still have Grok around to correct my grammar

No idea. My initial guess was because I was running long-running Fable 5 tasks related to healthcare, and I think that's within a specific classification of usage, but then why wouldn't they have suspended me earlier -- it doesn't make sense

I was paying for subscriptions and $20-plus one-off apps that each did some of these things — one to keep the Mac awake, another to lock or sleep it, another to force-quit apps. So I said f it, had Claude put together my most used actions from those apps into a menubar utility, and here we are.

> The article is confusing the architectural layers of AI coding agents. It's easy to add "cut/copy/paste" tools to the AI system if that shows improvement. This has nothing to do with LLM, it's in the layer on top.

I think we can't trivialize adding good cut/copy/paste tools though. It's not like we can just slap those tools on the topmost layer (ex, on Claude Code, Codex, or Roo) and it'll just work.

I think that a lot of reinforcement learning that LLM providers do on their coding models barely (if at all) steer towards that kind of tool use, so even if we implemented those tools on top of coding LLMs they probably would just splash and do nothing.

Adding cut/copy/paste probably requires a ton of very specific (and/or specialized) fine tuning with not a ton of data to train on -- think recordings of how humans use IDEs, keystrokes, commands issued, etc etc.

I'm guessing Cursor's Autocomplete model is the closest thing that can do something like this if they chose to, based on how they're training it.

In my defense, I wrote the blog post about quitting a good while after I've already quit cold turkey -- but you're spot on. :)

Especially when surrounded by people who swear LLMs can really be gamechanging on certain tasks, it's really hard to just keep doing things by hand (especially if you have the gut feeling that an LLM can probably do rote pretty well, based on past experience).

What kind of works for me now is what a colleague of mine calls "letting it write the leaf nodes in the code tree". So long as you take on the architecture, high level planning, schemas, and all the important bits that require thinking - chances are it can execute writing code successfully by following your idiot-proof blueprint. It's still a lot of toll and tedium, but perhaps still beats mechanical labor.

I clicked so fast expecting some sort of weird art experiment of maybe a hidden camera stuffed into a delicious cheese burger, as it goes through Munich.

Sadly, I was disappointed.

Hi! I wrote Doctor because I keep struggling with grounding on docs when working with agentic code editing (ex Roo, Claude Code).

Doctor uses crawl4ai to crawl websites, and then chunks and embeds them with langchain + litellm + openai, and finally stores all the vectors in duckdb. This allows your LLM to query the docs using semantic search over MCP, giving it grounded and up-to-date information for the things you're working on.

It requires an OpenAI key for the embedding process, but I'm working on giving users options in the future (different providers, local embedding using something like DPR or other transformer libs, etc.)

Interesting! Do you have an invite to spare? My email is in my bio

We had this same question when IDEs and autocomplete became a thing. We're still around today, just doing work that's a level harder :)

Hehe 12345678-b00b-4123-b00b-b00b13551234