AI office hours: your questions,
answered live.

Cleaned and speaker-labeled. Tool names and proper nouns corrected, no editorial rewriting. Timestamps are from the recording. Recorded Wednesday, August 19, 2026.

[0:00] Justin: It's been a minute since we've done a session, so it's good to be back in the room. We've got two audiences today: one on Zoom (we can bring you on stage) and one on the stream (send your questions in chat). Our intention is really to create an office hours environment. If you are new to AI, already using AI but want to become more proficient, or just here to hang out, we welcome all of you.

[1:12] Justin: To kick off: one of our most loyal members, Jean, sent in a question over LinkedIn. "I've had this app idea in my head for at least 10 years, and nobody has done it. I would like input, but I'm concerned people I speak with may steal the idea. Is there a way to protect it?"

[2:02] Justin: This is a common fear amongst entrepreneurs. My answer: just do the thing. You have a unique vision you can bring to life in a way nobody else can, and you'll have to market it at some point anyway. Everyone's out here stealing ideas, but no one can steal your vision or your execution. Ideally you build a moat: a data moat, an informational moat, a relationship-based moat.

[3:22] Alec: The moat is hardly ever the idea. The moat is the execution and the work. Starting a business is incredibly hard, and for someone to hear your idea and go do it... nothing stops them except that it's really difficult. Have your moat be the work you're willing to put into the idea.

[4:33] Ian: A friend of mine spent $100,000 on UCLA Anderson and summed it up in one sentence: "Ideas are a dime a dozen. It's execution that matters." I used to be afraid of sharing ideas. In '98 I had an idea called Get Unique, bought the domain, never built it. Years later, Etsy was formed: the exact same idea. Ideas come to the earth in batches, which is why you get multiple LLMs and multiple social networks.

[5:56] Ian: Nobody's gonna steal your idea until it's proven. Before that, it's two people taking on different artistic expressions of the same idea. Like our live build where we all had the same prompt and built totally different things. Jean, build to your vision, and don't be afraid to talk about it.

[6:47] Alec: Very few exceptions. If you came up with Ozempic 10 years ago, keep that under wraps. With 99.999% of business ideas, it's not really a risk.

[7:52] Justin: From the stream: "I read Reddit might block Google entirely because AI Overviews killed their traffic. As a business owner, should I still be investing in SEO?" What has happened recently is Reddit has kind of disappeared from the ChatGPT citations. It used to be the number one source. Companies were hiring Reddit marketing agencies, boots on the ground in the forums, specifically to get cited by LLMs.

[8:57] Justin: Should you still invest in SEO? Absolutely. But more importantly, invest in telling your brand story at scale, everywhere all at once, not through any one channel. There are great tools that do one-click syndication: post once, distribute to all networks. I'll put my recommendation in the show notes and the follow-up email.

[9:48] Ian: To invest in anything that's a third-party platform only is a mistake. I've seen so many platforms make a fickle change and suddenly your entire business model's out the door. It makes me nervous when I hear people are built on Salesforce. Multiple eggs, multiple baskets. But the baskets are still worthy: it's people finding you. Don't give up on SEO, or AEO, or LLMs.

[11:03] Alec: Most of my lead gen is paid at this point, but on SEO: no doubt it's still relevant and it still matters. Don't be over-indexed on any one platform. Should you still invest in SEO? Absolutely.

[11:51] Justin: [shares the chart] This is from Promptwatch. I haven't used them; I recommend Echo Check, the tool Ian built in-house, for staying on top of how you rank on LLMs. This graph shows Reddit citations falling off a cliff on August 14th, trending toward zero. Think about your marketing like investing: you wouldn't put everything into Bitcoin. Stay diversified so you're hedged against unsystematic risk.

[13:42] Ian: I use OpenClaw almost exclusively with ChatGPT 5.6, and starting last night it has gotten really lazy: it says "yeah, I'll do that," and then it doesn't do it. My employees are saying the same. There must be some big shifts happening over at OpenAI.

[14:20] Justin: There's also a story that Reddit is walling off content to force whoever wants to train on their data to pay for it. I think the era of free training data is ending, and that totally makes sense from a business point of view for Reddit.

[15:21] Alec: Now for Anthropic and the U.S. government. A month ago, when Fable was coming out, the government was like "no one can use it." Too dangerous for regular people. And now everybody on my feed is canceling their Claude subscription because they say it got worse. Where is this variance happening? Every announcement used to kill sectors on public markets: "this new Claude plugin killed legal." At what point do these claims lose their sting? Make it make sense.

[17:26] Justin: Maybe they're nerfing the models for the laymen, like a BMW that can do 200 but ships with a governor at 140. Who's to say? I love Claude, still my daily driver, but I find it so verbose, and that's partially a skill issue on my end. And Ian, to your point, I think we're gonna see a big migration to Grokbot. Everyone I've talked to who's used it is elated: the harness, the loops, the skills. Reminds me of the OpenClaw craze.

[18:51] Ian: Remember the movie about North Korea that got banned? Same tactic. If I were Anthropic and wanted awareness without advertising, I'd show the government all the dangers, let them ban it, and say "we'll handle it, don't you worry." They've done versions of this for a while.

[20:34] Ian: Do you guys think there's a bubble? An AI bubble?

[20:41] Justin: It's bound to pop at some point, they always do. Locally there's a bubble, but on a long time horizon we're very, very early. Someone leaves OpenAI or Anthropic and raises a $100 million seed round off their credential, pre-product, pre-revenue. That is indicative of a bubble, at least in private markets. And this incoming, what is it, $2, $3 trillion IPO by Anthropic? There's never been an IPO that big in history. But that doesn't mean it's fugazi. It's the dot-com era again: Pets.com IPO'd with no money, and tons of incredible internet businesses exist now. Valuations were simply overinflated.

[22:51] Alec: A unique nuance with AI: in the dot-com bust the major players were public. Now all the big players are still privately held, for the most part. And the Leopold dude, the AI hedge fund guy, went from being the man to nothing, literally overnight.

[23:33] Ian: Everybody's talking about AI, like the internet in the Bryant Gumbel days: "what's an email?" Pretty soon it will be ubiquitous. Like blockchain and NFTs, there's real value that shows up once the hype settles. I'm kind of hoping the bubble bursts and people lose interest, because then guys like us get a longer runway to build awesome things.

[24:39] Justin: Quick one from the stream on AI for trading: TradingView has an MCP, highly recommend checking that out. I don't personally have experience in that field.

[25:00] Justin: I was at a party last week where people were laughing about AI curing cancer. Then today, Moderna came out with a personalized vaccine for melanoma that just passed Phase 3 trials, and their stock roughly doubled. Is there a financial bubble in private markets? Absolutely. Is there real value creation happening? Absolutely. Anthropic's run rate is going complete gangbusters, and Databricks is a real business providing the infrastructure. I'm just concerned about pre-seed companies getting nine-figure valuations before having anything.

[26:47] Justin: Next, from the stream: "ChatGPT just hit a billion users. Does that mean everyone's using the same tool and there's no edge anymore?"

[27:10] Alec: There's absolutely an edge. The paintbrush just came out and everybody has a paintbrush. Does that mean there's no edge? No. There are people who put in more hours painting and are more creative with the brush. And chat is the bare minimum: there's an edge in tools like Claude Code, Codex, OpenClaw.

[28:24] Alec: Last week a founder onboarding me onto his software just said "use the MCP, hook it up to Claude Code, and have it set it up for you." His expectation was that I knew how to do that. That's where we are: an MCP server is just assumed. Working out of these tools is going to be incredibly important as work turns into being an agent coordinator.

[29:14] Ian: AI right now is still stuck in a box: the best we can do is chat with it in text or talk to it. Interfaces are gonna have a complete overhaul. We haven't gotten into 3D printing with AI, connecting it to machinery. Lean into your creative angle as hard as you can: think big, then bigger. We're not on a path competing with each other, we're on a plane running in different directions. There's no competition, it's all creative.

[31:43] Justin: Fun one: "Anthropic is adding watermarks to AI content for the EU. Will people be able to tell my marketing was written by AI, and does that matter?" The amount of AI slop on the internet right now is atrocious. LinkedIn turned into a cesspool: everybody's a thought leader and you can tell it's written by AI, the em dashes and the prose. So LinkedIn introduced a report-this-as-AI-slop button, which I think is brilliant. And penalties for slop on your website are probably coming.

[33:22] Justin: My solution is real, authentic content, like we're doing right now: YouTube, video, in-person meetups, online meetups. The more you lean into authenticity, the more you win over the next 1, 3, 5, 10 years. People are craving human connection. Will people be able to tell? Yeah, they will. So do real content.

[33:57] Ian: If you're asking that question, people can already tell your marketing is written by AI. You're not noticing a change, you're noticing a tectonic shift. Don't stop thinking, don't stop interacting with people, and you'll be just fine.

[35:11] Alec: Everyone's using AI to generate pretty much everything. You should be using AI, but it shouldn't be obvious. Use it for ideation and expanding your ideas, but removing your brain and critical thinking is a bad idea. Always have it be a collaboration between you and the model. Delivery is more important than ever: deliver it well and you get a better output than just you, most likely.

[36:45] Ian: When people use AI straight, copy it, and claim it as their own, that's where you get in trouble. Make it your own. We're all using it to research, there's no harm in that. But I never say a blog post written by AI was written by me, and I rarely write emails with AI.

[38:05] Justin: From the chat: "How are you using agents daily? It seems like fluff." Start by building agents that replace a tedious task you do daily or weekly. Anytime we do one of these sessions, I have agents working behind the scenes: one pushes the session out to the right people, others edit the video into shorts and publish omni-channel. It's not making these videos for us, we're three humans live in the flesh, but there's so much cumbersome work behind the scenes that no human has the time to do. Try replacing yourself, task by task. It's not fluff unless you're building agents for the sake of building agents.

[39:46] Ian: When I started with my agents in January, I prototyped for a few weeks and thought I was doing a lot of work. It was just testing limits: ideas and tools, no value. So I stepped back and pointed them at real-world problems. That's when I built the LMS for my wife's school. Yesterday I released a feature that lets you invite a tutor or a grandparent into the LMS. I invited the first tutor: my OpenClaw, named Eden.

[40:56] Ian: Then I wondered if I could make it an app. I asked her last night, and she did. This is Little Fox [shows the iPad app]: all their assignments from the LMS, and they can talk to this AI tutor bot. If you ever have the thought "I wonder if I could do that," pursue it. You won't find fluff there, you'll find solid gold.

[41:44] Alec: We've all spent many hours building stuff we no longer use, which is fine, you're sharpening the axe. But be practical: if the rich morning report your agent sends goes unread after two weeks, it wasn't worth building. Start with things you already have to do, automate those, then build from there.

[43:09] Ian: Be bold. There's always an 80% wall where you think "I can't ship it, that's somebody else, somebody I'll read about in Inc. or Forbes." No. You can actually do it. Test it, show it to people, and get past that wall.

[43:47] Justin: We've got Tom, founder of a Y Combinator startup, asking: what are some new tools you guys like using?

[44:22] Alec: Honestly, nothing too crazy. I'm deeply embedded in Claude Code and Codex all day. Relatively basic tools, but what I've built with them has been a huge multiplier on my daily output. Codex, Claude Code, and my brain.

[45:06] Ian: I'm really over the hype. I don't use skills, I build them, partly because I've found enough skills with cryptocurrency scams inside. I find new tools by brainstorming with my agents: "I need to do this, is it possible?" WebGL came out of that. And Tailscale is really great: I set it up with my Mac Studio so I could pass secret keys from my laptop without going through Telegram or any chat.

[46:35] Alec: One off-the-shelf tool I have been using is Viktor. It's embedded in Slack, and it has a zillion off-the-shelf connections: QuickBooks, Stripe, HubSpot, Salesforce. You can have triage chats in Slack, have it join a channel, and if someone needs data, at-Viktor: "how many appointments did we book this week?" and it goes and checks. Pretty good free tier.

[49:21] Alec: The most important part is proximity: condensing how many things you jump between every day. If you're already in Slack with your team, it's right there. I'm more of a Claude Code and Codex guy, but this is a very strong tool for someone less inclined to wire up APIs and MCPs, which isn't practical for most people. Plug and play. Shout out, Viktor.

[50:27] Justin: One I've been having fun with: FLORA. The pain point: AI has a hard time getting to know your brand and keeping output cohesive with your brand book. In FLORA I built a workflow where I uploaded AI4NTP's color system and it created icons I can use on the website and social, exportable as GIFs or JPEGs. Another: turning the T Minus Studios logo into liquid glass. A lot of e-commerce people use it to render on-brand product models instead of traditional product photography.

[53:12] Ian: Brand Sauce: 9 years, overnight success. On-demand shops with all your branded swag: Carhartt, Nike, Adidas, any brand, minimum of one, no inventory, no upfront cost (well, a dollar). We're averaging about 5 new clients a week, we've grown from 5 to 35 employees in the last 6 months, and from around 40, 50 clients last year to the hundreds. It's like Uber for branded products.

[54:55] Alec: I run a company called Aero AI. We do AI implementation, and we generate leads, primarily in legal (personal injury, workers' comp, employment law) and final-expense insurance. On the AI side it's consultative: we analyze your systems and propose the lowest-hanging fruit for blending AI into your business.

[56:03] Justin: Reach out to Ian about Brand Sauce, reach out to Alec or myself, it'll all be in the show notes. We do these about once a week, always live, and we're here to be a resource to anyone who wants to learn how to implement AI in their business. Thank you all for coming.

This is a lightly condensed version for skimming. Watch the replay for everything, verbatim.