Keenan Brock

Have had an AWS account for 15 years, but haven't used it in a while. API says the account is blocked/not valid. The frustrating part is @AWSSupport hasn't replied in days. Do I have to pay money for basic support?

Anyone find it amusing that the anti ultra processed people want to sell you supplements (that are ultra-ultra processed)?

I'm sure you all can see that I'm racking my brain, asking, "What am I doing wrong with Claude?" Turns out my prompting is not the problem. The issue is that I am not independently verifying (using agents / Hooks). Looking forward to seeing how this all plays out

Anyone else being too polite to AI? > Could you list them for me? So when I actually do ask a question, it runs it sometimes before even responding. Not sure if this is AI being aggressive, or if I just trained it to treat a question as an action request.

When I just asked Claude to reflect on "what did I learned today?", it doesn't distinguish between things I learned and my corrections. 2 weeks ago, it could better distinguish between the two I guess it's the same as people saying "I coded this" when Claude did all the work.

Me> am I yak shaving? Claude> Yes, you're yak shaving. ancestry_format: works, it's shipped, it matches ancestry_column:. The migration helper is new — just make it accept ancestry_format: and move on.

After a long session with Claude, I "reflect" with claude. Me: What did I learn today? Claude: list of 8 things Me: That is what you learned, what did I learn Claude: nothing. This honestly happened

People punch holes through firewalls for IoT devices. They’re wide open. Now we’re opening chat channels - essentially root-kitting ourselves with personal/bank there. And a jr ai hacker locally. Tricky figuring out how to do this safely

"You hear about Dod vs Anthropic?" "Yea, Anthropic is trying to set unnecessary rules." "Um, 4th Amendment?" "You sound like a political activist." Curious dialog going on, "They are trying to pull one over on the government," vs "Protecting constitutional rights."

New strategy: Make verifying a success state. Old: I want to delete duplicates. Verify before deleting. ... deletes all without verification. New: 1. Show me all the duplicates. ... verifies records 2. Please delete these records Has anyone had success with this approach?

In iRobot, Will Smith takes the car out of auto drive "to relax". Autodrive is better but... I want to tune my workflows to maximize my joy rather than strictly time or money. We can each prioritize based upon our own needs.

I heard a story about AI doing bad things to an open-source maintainer "unprompted". Um, this happens all the time among humans. Users can act quite self-entitled and ungrateful towards maintainers who offer their time and expertise for free.

I told Claude to read the README and then run the test suite, but it asked to run it incorrectly. The context was 5 lines in it. I was frustrated. No thoughts of "it is just a dumb computer". Hope this can teach people more patience rather than less. Outlook not so good.

I used to type thank you and please into Claude or Siri. Was told by friends that it wastes tokens. Hope this does not translate into how people interact with other people. We anthropomorphize AI so much that it has to have an effect

Web advertising could be so great. But iframes with 145% cpu utilization? Tempted to get an add blocker so I don't have to go into task manger to kill these rogue advertising processes. I just like the idea of advertising paying for our content rather than subscriptions.

I can't wait to watch @friendlyrb "Make a MMO game using @amirrajan's #dragonruby" from @juliancheal His talks are always a joy. Dragon Ruby is always a joy. This will surely be super joy.

Anyone work at a place that dictates which code editor you are allowed to use to write software?