Some Of The Things #43
Welcome to the 43rd entry of my public journal, where I share *some of the things* I've been exploring across work, tech, wellness, and life.
Building a Farcaster AI Agent
I decided to take a stab at building a Farcaster AI agent trained on the hundreds of hours of video content NounishProf and I have recorded for GM Farcaster. People have commented that our video archive is like a time capsule of Farcaster history, but we haven’t been able to leverage it yet because video is not inherently searchable the way text is.
I always knew eventually we’d have an LLM trained on our content, but I hadn’t done anything about yet it, mainly because I didn’t know where to start, what steps were involved, and how hard it would be. Seeing recent AI agents like Aether, mferGPT, AskGina.eth and Clanker explode on Farcaster only reminded me it was something that had to be done sooner or later. Then two weeks ago I saw Six’s comment about Aether:
It's true that anybody *can* make a bot/ai agent for any community - the tech is all out there in the open.
But anybody *can* do a lot of things, so much so that the statement becomes meaningless. Out of the set of people who can do something, only a few people /actually/ do it, and within that set, even fewer people actually do it in a way that is valuable.”
And I decided I had to stop waiting and just do it.
With the help of some developers who have done similar projects and ChatGPT, I’ve been making my way through the steps to get to an MVP.
Will I be successful in seeing this project through and actually making a bot that is useful and functional? I don’t know. It’s certainly possible but not gauranteed. But there are two things I do know to be true: 1.) I can try anything, and 2.) I can learn anything. So after a week of playing around, here are some reflections:
AI coding assistants / copilots are insane. This isn’t news to anyone but to experience it personally is still mind blowing.
Training an AI on a dataset is harder than I expected, I started with transcripts from just 2 videos and the initial result is laughably bad. ChatGPT makes everything seem so easy so I was surprised at how bad my first iteration is. I think I’m realizing the work involved in training an LLM is classic schlep.
I forgot how much fun coding is. It’s my ideal flow state and the pace of learning is so fast. I’m waking in the middle of the night with dreams about code.
I’m not getting fast enough feedback loops. I’m working on the thinnest possible vertical slice, just so I can see something end to end. Then I can figure out what needs to be optimized. When I tested my first embeddings and realized I wasn’t happy with the result, I was tempted to improve them but I decided to let it be and I’ll come back to it later. I want to see a full end to end working flow before I decide what should be fixed.
The internet is still goated. As I was doing research I came across a recently published blog that more or less describes what I’m trying to do: Podcast RAG. The internet allows us to learn anything, thanks to people who work in public and share their work.
How’s the job market?
I caught up with a friend last week who has been doing tech recruiting since the late 90s. We get on the phone once or twice a year to talk about hiring trends and best practices. For example, when we spoke during the great resignation in 2022, we touched on a shift towards more flexibility (hybrid and remote work), the dangers of drawing from too-narrow or too-wide a candidate pool, the need for speed in a hot job market, and the reasons people enter the job market (it wasn’t compensation).
I asked him how the job market is in 2024.
Hiring Trends 2024
Volume is down but not as bad as it was 6 or 12 months ago. In June of this year he was recruiting for just six active roles, much less than his usual sweet spot of 15-20. By August things were picking up and he was back up to 10-13 roles, but hiring is still not back where it used to be.
Hiring is a muscle and will weaken if not used. Investment outflows are increasing and money is starting to move again which usually translates to more hiring. But hiring is a muscle, and it atrophies when not used. He said it almost feels as though companies have forgotten how to hire and he said he observes this trend particularly in companies where IT reports to the CFO. CFOs are struggling to forecast with confidence which is putting pressure on hiring. The CFO will grant budget but IT “better be sure they need it.” This pressure is making IT managers hesitant to pull the trigger on new hires, even when they need it.
EQ Skills are In Demand. I asked him what skills companies are prioritizing becuase I’m always curious what technologies are popular. He’s seeing a growing emphasis on emotional intelligence (EQ) over technical skills. He shared an example of a company that recently took a big risk by hiring a young candidate with almost no experience, betting on potential rather than a proven track record. I imagine AI copilots will only further this trend.
Some of the (other) things
In my last update, I wrote about a camera I pre-ordered that is designed to prevent deep fakes by using cryptographic proofs, but I didn’t mention the name of the camera. It’s called the Roc Camera and the company behind it is called Faust.
McKinsey’s 18 next big arenas of competition. “Arenas earn far greater profits than other industries do, they spawn a disproportionate number of global giants, and they offer unusually strong opportunities for new entrants to become powerhouses.”
I saw a screening of The Last Showgirl last weekend, followed by a Q&A with lead actress Pamela Anderson, director Gia Coppola (of that family yes), and writer Kate Gersten. The film follows Anderson’s character as her long-running Vegas show faces its last two weeks, forcing her to look back on choices and relationships and figure out what’s next for a 57 year old woman who, as beautiful and strong as she is, is not what the new shows are looking for. It reminded me of a recent discussion I had with my Four Old College Friends about how young women often rely on their looks, a depleting resource over time. My wish for young women is for them is to focus on skills that compound over time—education, business, investing, technology — and avoid focusing too heavily on looks.
Some of the pics
Romanesco and cauliflower at the farmers market;
Leaves under water, so still. A zoomed out photo of the same scene.
The Q&A panel after The Last Showgirl
A Note to My Readers
I started this substack in December 2022 as an experiment to see if developing a writing habit would help clarify my thinking and/or provide other benefits. You can read about my original intentions in my first post or my more recent reflections after sticking with it for a year.
I write about twice a month and share musings, meditations, and links to things I’m finding interesting as I build out a farcaster-native media company, a modern technology consulting company, raise my kids, and have fun creating and learning in the worlds of crypto, tech, finance, science and wellness.
Thank you for supporting my writing and journey. If you’d like to get in touch you can reply to me here or find me on X and farcaster.
Until next time, keep putting good into the world. —adrienne🌏❤️