Transcript

[00:00:00 Intro Music]

Life's got flavor,
got style, got zest.
It’s Jon and Lemon,
we’re feeling our best.
Talking shop, we’re taking it slow,
crafting life as we…

[00:00:52 Kicking off after the hiatus]

Stan: Man, I don’t even know where to begin. Are we doing this like a real episode? Why not. Maybe we can use it. Hold on, let’s talk this out. When was the last time we did this?
Jon: Going into 2022. December 2021, talking about the new year. So, almost four years. A lot has changed.
Stan: I don’t think that much has changed.
Jon: If you say so.
Stan: The website’s back up. People might be seeing the feed in their apps, and hopefully files download fine. That remains to be seen. And I guess we have our AI overlords to thank.
Jon: Absolutely. I don’t think we would have gotten the site back up without them.

[00:02:07 AI made hard things easier]

Stan: AI is only as good as the person driving it. Maybe that’s a good place to start. AI brought us to this moment because it made things that were hard easy, or at least easier. That might be the biggest change in four years.
Jon: I’d say probably, globally speaking.
Stan: Individually, a lot has changed. Some problems went away, some multiplied. When we started this show, the idea was to take the conversations we had on the phone and do them on a podcast, especially the new tech talk.

[00:03:10 Marvel, milkshakes, and cruises]

Jon: We talked about a lot. There was that epic milkshake run. Everybody got sick of us talking about Marvel, and I don’t think I’ve seen any Marvel shows or movies since then.
Stan: I’ve stayed up to date, and I am less happy for it. We also talked about vacationing. I think we teased a cruise, which we did in early 2022.
Jon: I think so.
Stan: I’ve quite literally sailed around the world since then.
Jon: You unleashed a beast.

[00:04:03 Using AI for resumes]

Stan: You were telling me you’re going to have GPT help rebuild a master resume from the ground up. That’s interesting because I tried this last year. I used to say, make the resume an annual exercise. When I left Salesforce, I had not kept mine up to date for eight and a half years. Starting from that in 2021 was insane.
It’s hard to write honestly about yourself and tailor a resume to a job description. In 2025 you do not get one resume to rule them all. Robots can help.

[00:07:03 What Jon tried and why it went stale]

Jon: I had a custom GPT acting like a strategic resume builder. I fed it resume versions and job postings. It tailored content. It worked at first, then degraded as it got out of date.

[00:08:00 What a custom GPT actually is]

Stan: We should clarify “custom GPT.”
Jon: You layer knowledge and instructions on top of the base model so you don’t retype them every time. If you want consistent structure, it helps. Think a recipe builder that always outputs in your preferred format.
Stan: Projects also let you set an overall prompt across chats. The line between that and a custom GPT isn’t huge. The big differences are shareability and reuse.
Jon: More important than either is setting custom instructions in settings. That’s how you get from generic to tailored.

[00:15:10 Voice mode and workflow upgrades]

Stan: In February I tried speaking to ChatGPT in voice mode, describing what I’d done in each role, asking for a long-form work history. Two years ago I recorded Zoom, exported a transcript, then summarized with an LLM. Now I can mostly talk directly and get usable prose. Models got better. I generated a few resume versions. I want to add to the long doc and let today’s models carve out what I need.

[00:18:18 Master resume strategy]

Jon: A CV is everything. I don’t need that, but I like a master doc that tells my work story.
Stan: Hot take. The master doc doesn’t need to describe impact. Capture exhaustive facts. Let the LLM derive impact and tie to outcomes, while you keep it honest.
Jon: What about companies that no longer exist?
Stan: Capture facts, avoid hallucinations, and fact check.

[00:22:31 When models get “cute”]

Stan: I asked Claude about running shoes. I said I was old and slow, wearing HOKA Bondi 9s, considering Brooks. It titled the chat “Brooks Running Shoes for Seniors.” A bit much. It apologized. Point is, steer it and fact check.

[00:25:10 Fresh run for the annual refresh]

Jon: For this refresh I’m not using a custom GPT. I want a clean slate. My first prompt was: “Let’s rebuild my master resume from the ground up. You’re a recruiter with a 90 percent placement rate in high-tech SaaS. You know what questions to ask and how to craft a standout resume.” It proposed a plan with seven foundation questions, then structure, summaries, impact bullets, metrics, and an ATS pass. I’ll go one question at a time and force follow-ups.
Stan: Defining the persona in your prompt helps. Also, check out Darmesh’s Meta Prompt to improve prompts before you use them.

[00:29:00 Human creativity still matters]

Jon: Human creativity still drives quality. AI isn’t magic. Give direction, challenge it, correct it, tell it to remember stylistic preferences, and it adapts.
Stan: Most people don’t know how to get there yet. Our backgrounds with scripts and automations help. Adoption will spread beyond engineering.

[00:31:20 Who adopts fastest]

Stan: I see an odd demographic split. Older engineers and very young engineers are more receptive. The middle cohort that lived through long stretches of slow tool change is more skeptical. Rapid OS transitions early on, then long stability like Windows XP shapes how people handle change.

[00:37:49 Business software cadence]

Jon: Where does business software fit?
Stan: SaaS exploded, but enterprise adoption stayed measured. At Salesforce we shipped three times a year to avoid whiplash. We had a long slowdown, and only recently does it feel like the pace picked back up.

[00:39:17 Subscriptions everywhere]

Stan: Subscriptions are another sign. Two years ago I had maybe One Password. Now I track multiple, and AI added more. I’m deciding whether to cancel Claude, ChatGPT, or both. Add Grammarly, which is basically an AI product now.
Jon: Do you still need Grammarly?
Stan: The UI is useful. It fixes a lot of sins from LLM drafts. Apple’s tools can be heavy handed. Grammarly is shifting identity after acquisitions, so we’ll see.

[00:41:11 Cheap iteration and experiments]

Jon: Rebuilding the site, experiments are cheap. The robot opens a PR, we review, keep or toss. Usage caps can bite, but trying and discarding is great.
Stan: Exactly. For your resume, even if you toss this run, getting a solid neutral work history is a win.

[00:42:18 Coding session idea]

Stan: I’m thinking about a YouTube vibe-code session with Claude Code. Vanilla project, start to finish, time-boxed to an hour to show the process. Even if live is boring, the recording could help.
Jon: Try it, then run the video through an LLM for process critique.

[00:44:01 Video understanding and Notebook LM]

Stan: If I can find one that handles video well, I’m in. Conversation mode can analyze short clips but it still whiffs sometimes.
Jon: Notebook LM might help with YouTube videos.
Stan: Good call. It can ingest PDFs, sites, YouTube, audio, and more. Worth a try. I also installed Perplexity Comet, then ran out of time. Classic.

[00:46:02 Wrap]

Stan: If the recording isn’t trash and the new stack behaved, maybe we have our first episode. We’re back.
Jon: We’re back.
Stan: Till next time.

[00:46:15 Outro Music]

That’s a wrap, yo, time to unplug.
Stan and Jon just served that mug.
Twist of Lemon, sharp and clean,
truth and tech on your podcast screen.
Catch us next time, do not let it slip,
wisdom packed in every clip.
From faith to fiber, finance to code,
we are taking the mic down the open road.
Stay sharp, stay bold, that is the theme,
with a lemon twist and a podcast stream.