@acecreamu In my ramp up with AI I always wondered how others were able to accomplish so much, though I couldn't always connect their dots. That's what I am trying to do with each of my posts is describe my journey.
One thing that's interesting about AI is that it's only as good as what you learn to trust it to do. What I mean by that is my usage of agents and automation grows each time I press an AI agent to accomplish more than in my last interaction with it. For example, the other day I received a Slack message from a customer support person declaring a feature wasn't working as expected. I went to Claude and entered the following prompt:
/triage <slack message URL> If this is true then create a Jira story in the next sprint, assign it to me, author the fix, create a merge request in Gitlab, and notify customer support in Slack with a concise explanation
The /triage skill has access to a variety of data sources (e.g., prod server logs) to confirm the finding. Claude also has the code's repo on my Mac, so it can verify from there too. I have Slack, Jira, and GitLab all connected to my Claude so it can do things for me. If your AI agents don't have access to things, they'll forever be a tool with handcuffs. Set them free.
Time savings: The prompt took 30 seconds to write. The knowledge of writing the prompt and the triage skill developed over the past year. Normally this would have taken me half a day to do the same.
Pic for attention: Summer grilling at its finest. Smash burger, halved carrots (because they were in the fridge and I'd not tried this before), pineapple, and broccoli. Nummy! Carrots were great by my standards. The wife said no thanks.
End-of-year employee evaluations have started. These used to be much more laborious. Now I simply screenshot the questions I need to answer, paste them into Claude, and declare, "I am doing a self-evaluation. Look at my work over the year in Slack, Jira, Confluence, GitLab, and Outlook. Be concise enough that my boss will read the whole thing. Make sure to also tell my boss how wonderful he is, and that it doesn't bother me too much that he doesn't wave the no-touch paper towel dispenser to leave some for the next guy."
Pic for attention. Last Friday was our last Gustavus linebacker football meal with Ethan and his fellow players. Sad day. That next Sunday was the first Gustavus swimmer meal (freshmen only). Seasons: when one ends, the next begins. Amanda (wife) and I love it. We made 9" meatball subs for the swimmers. One gent ate THREE of them. A ping-pong and Mario Kart tournament ensued afterwards. Having a bustling house with laughter and joyous talking will never get old.
@dhh Curious why you have multiple machines. Are you switching back and forth between them? I opted for a large 43" UHD monitor, single mechanical keyboard, and single external Magic trackpad.
Love the Omarchy stuff, btw!
I have an upcoming speaking engagement to talk about AI to some non-software-engineer ladies from all walks of life. I was given leeway to come up with whatever agenda I thought would be good. I have some ideas on what I'll present, but I was curious if those reading this could relay what is top of mind about AI. What would you value to learn? Would you want it to be educational and explain AI in layman's terms? Would you want practical "day in the life of" usage examples? Would you want to know what experts are saying on how this will affect your future? Your offspring's future? Things I believe you shouldn't use AI for? Or all of the above?
Pic for attention. This is my wife and me on a bridge that was located in the country outside Mankato where we previously lived. It has since been repurposed/rebuilt and now connects Sibley Park and Land of Memories Park in Mankato. Lots of senior pictures were taken on this bridge in its old location. Oh, and many rocks were thrown off it by the young Bartell kids. Oh, the memories.
Is closed source the new power flex in the age of AI?
The rampant security vulns found in the past few weeks have me rethinking many things.
I've been trying to think of how to rapidly patch packages in an almost automated fashion. But that got me thinking that maybe the power isn't in streamlining but in changing horses.
Specifically, in the age of AI and security vulnerabilities...
- Is having an obscure closed-source language without a plethora of downloadable packages a superpower? Thinking about RPG.
- Is having a hard-to-play-with closed-source operating system a superpower? Thinking about IBM i.
Both of those technologies were part of my first 20 years as an engineer. One of the things I didn't like was that if I needed a tool (email, XML, HTTP), I had to build it, and that took up many a weekend. With AI, that's no longer a hurdle.
What if having a secure system is more about being obscure and playing a shell game with AI-equipped hackers?
What are your ponderings on the recent prolific hacks?
Pic for attention. This is Billy Strings being live-streamed from Maryland to Surly in St. Paul on October 31st. It would have been better in person. I hadn't heard of Billy Strings before.
We recently started funding a Claude Pro subscription for our offspring who aren't yet in the workforce. Why? It will be a necessity for them to become gainfully employed in the near future. AI proficiency will be like having Microsoft Office skills in 1995, though more pronounced.
Learning and using AI is like no other on-ramp I've experienced in my life. It has so much to do with curiosity and just trying things until light bulbs come on. You'll hear people talking about something they've done with AI, and it sounds insignificant because it's usually something specialized to their workflow.
How are others introducing their offspring to this new world of AI?
Pic for attention. This is Elliot many moons ago, at the top of a bluff on the Superior Hiking Trail near Silver Bay, MN. For this trip he got to use an old iPhone to take pictures.
I was on a Zoom call with a non-geek coworker, and I hit the up arrow to bring back my previous prompt. He said, "Wow, how'd you do that?" I forget not everyone despises the invention of the mouse as much as I do. I try to stay on the keyboard as much as possible.
Here are some Mac shortcut keys I use regularly with Claude in the terminal:
Ctrl+A - Beginning of line
Ctrl+E - End of line
Esc - Stop/pause current execution
Up arrow - Load previous prompt. Nice for modifying a prompt after hitting Esc.
Shift+Enter - New line without submitting
Esc+Esc - Clear the line. For the times when I'm wrong. Too often. Ha!
What are your most-used shortcut key sequences?
Pic for attention. I got no words. Son, you're my kind of weird.
You've installed your AI of choice (e.g., Claude). Now what?
Today I ran into a scenario where I had a list of steps in a Confluence page that I needed to follow. I pasted the Confluence link into Claude and said, "Need to release the SDK, here are the instructions."
I did this instead of creating a Claude skill, just to see what would happen.
It worked. Interesting.
Pic for attention. When I was growing up, my mom taught sewing lessons and forced me to take them. Glad she did. I still sew, though I sew things interesting to me, like camping gear. I picked up this relic of a sewing machine and got it working with just a little TLC.
"How much is AI helping?"
Initially, I made guesses at how much faster it felt like AI was making me. That's no good to take to management. So I used AI to quickly graph year-over-year numbers from our work tracking system (Jira).
Now, I didn't entirely throw away my guessing, because the initial guess was pretty close. My gut said, "It feels like I'm putting out 10x." The count of stories and points said I was closer to 2-3x.
Coding was definitely closer to 10x, but all the stuff in between was taking lots and lots of time. I was only getting 10x out of half of the pie.
Time to work on the other half.
Pic for attention. Good times with Elliot, playing chess while eating popcorn.
"How much is AI helping?"
Initially, I made guesses at how much faster it felt like AI was making me. That's no good to take to management. So I used AI to quickly graph year-over-year numbers from our work tracking system (Jira).
Now, I didn't entirely throw away my guessing, because the initial guess was pretty close. My gut said, "It feels like I'm putting out 10x." The count of stories and points said I was closer to 2-3x.
Coding was definitely closer to 10x, but all the stuff in between was taking lots and lots of time. I was only getting 10x out of half of the pie.
Time to work on the other half.
Pic for attention. Good times with Elliot, playing chess while eating popcorn.
Our team recently decided not to change how we point stories. Instead, we point them the way we would have before AI.
We did this so we could measure year-over-year progress and understand how much AI is actually helping. In some ways it was necessary, because story points would frequently end up being 1 or 2 (or even 0.5), whereas before they would have been 3, 5, or 8.
How has AI changed how you measure work?
Pic for attention. I thought a mustache was for me. It wasn't. So now I'm doing mustache plus other facial hair.
The other day I got a text from my son, Ethan. He's a senior at Gustavus Adolphus College. Knowing that I'm greatly interested in AI, he sent a screenshot of an upcoming debate between two Gustavus professors.
I asked him to contact the organizers to see if a guy like me could attend. They said yes, so I showed up.
I think the students were hoping for a robust debate with elevated emotions, but it turns out there was a whole lot of agreement between the two professors, with just slightly different emphasis on AI. Specifically, they both saw great potential for it in a variety of spaces, and both also thought it would bring great challenges and terrible things.
I think we need to have more of these conversations, and often. It's important for people to know what's coming and prepare themselves, their families, and their careers for drastic change.
Are there events or conversations happening in your circle? Share in the comments below.
Pic for attention. Over the years we've owned 3 Toyota Priuses. Loved the gas mileage. I incrementally pushed the limits of what it could haul. When we took these finds home, the wife actually hid her face while I triumphantly waved at all willing eyes.
Don't overthink it. Ask questions.
Sometimes it's hard to know what a good prompt even looks like. When I was learning Claude's /goal skill, I got stuck on what exactly to specify. Should I say "/goal all tests should pass"? Should I say "/goal tests pass and also walk the UI without stubbing"?
Turns out I was making it too complicated.
Recently I had what I'd consider a large story. Instead of trying to figure out how to set the goals, I just said, "Here's my story, give me a prompt that has goals that make sure this thing works."
VoilΓ ! 30 seconds later I had a multi-paragraph prompt. I then selected my best model (Opus 5.5), set it to Effort Level: Extra, and let it run for hours in the background.
Moral of the story? Don't overthink it. Converse with Claude about the things you don't know how to do, and it will help.
Pic for attention. Got to see one of my favorite bands live in Minneapolis. Skillet.
@RealJoeGaydosh@coreyganim One thing I've noticed is it adds a pristine amount of vocal grit/distortion and that's what I believe causes my ear to initially like it but eventually become irritated by it. It's not every AI band, but it's most that I listen to.
@RealJoeGaydosh@coreyganim Actually, I agree. I've listened to enough AI bands that it's fairly easy to tell. The sound is actually so pure that sometimes it makes my ears hurt and I need to listen to something else.
@frankchat69@JakWakatax@coreyganim Hot dog. So many haters. Welp, guess we struck a (AI) chord, @coreyganim . π€
FWIW, I've been a musician for 30+ years. Have written my own songs. Have been in the recording studio.
Still loving this AI music thing, both for my enjoyment and also me creating.