MrAssisted

What if a "product" is just a config file and a directory of domain skills?

my custom AI agent harness is really coming along, I'm gonna get so much work done

"We'll subsidize your inference, but only if you sit here and waste your time doing it manually instead of having your agents do it" Please let this be a short-lived era.

so much obvious low hanging fruit to make the major coding agents more agentic. The labs have to be sandbagging at this point

I got neverending video generation working on 3x 5090s

I got to the future but I had to drop to 360p 😅 also B200s are expensive :/

.

providers make it so hard to tell what ai inference costs you basically have to hit another llm afterwards with some context and have it guess to the fraction of a cent

it's dangerous not to --dangerously-skip-permissions

still babysitting, still (barely) useful as a human. how long do I have left?

It would be interesting if the cycles are so compressed at this point the `AI winter` doesn't end up being meaningfully distinct from the `AI boom`, they just kind of play out superimposed, and you see one or the other depending on how you squint

revisiting eloquent javascript, a relic from a different age

OHHHHHH dang cursor shoved claude code in the overflow menu THAT'S COLD

I'm sitting with 3 current major USA university students, 2 comp sci 1 comp eng: none have tried cursor or claude code none of them have heard of attention is all you need none have heard of arxiv wtf

if you're not absolutely blasting Three 6 Mafia while trying to get serverless cloud ai vtubing working right now I can't relate www.youtube.com/watch?v=afO7...

heard github copilot added gemini 2.5 so I paid them the 10 bucks since that would be way more cost effective than chugging through $30/day in pro requests in cursor Copilot did add gemini 2.5, but it's not enabled in agent mode so I literally don't care. Word to the wise.

couldn't figure out why my serverless endpoints were spinning up for no reason when there were no bugs in my code that would cause that turns out I had actual users

any time I need to containerize some AI thing it means minimum two full days of dev time tied to my 3090 desktop pulling my hair out there has to be a better way and there are tons of product opportunities here but I can't envision the solution due to the constraints

please release your docs as a giant .md file so I can just paste it into gemini 2.5 along with my prompt

gemini 2.5 is so good that cursor's job now is writing shell scripts that log all the context from my codebase relevant to copy paste into giant prompts of entire new features for 2.5 to one-shot build It's the ultimate "I'm sorry, I'm seeing someone else" of vibe coding

I want cursor's UI and pricing with claude code's smarts and ambition

You can now link a youtube video to google gemini AI to include the content of the video in your query!

The future of at least some portion of frontend is just rendering HTML the LLM spits out. Like i2i-realtime for the dom.

My new blogging framework is telling cursor and sonnet 3.7 to add a blog to my backend and having the entire thing work perfectly in one shot

I actually have a whole other SaaS PricingLink.com I built alongside AIRanksEverything Completely different target market Should I make separate youtubes/socials for making content around each product or just post everything as MrAssisted?