THE NEW BUILDER

Jesse Allouf is the chief marketing officer of Tire Agent, one of the largest tire e-commerce sites in the country. He’s also one of the most advanced AI users Brian knows personally.

What they cover: what Tire Agent actually does and why tire selection online is harder than it looks.

Where AI touches the customer experience today, and where Jesse keeps it out on purpose, because a wrong tire recommendation is a safety problem, not just a bad review. Why Meta and Google already run AI on every advertiser’s behalf whether you opt in or not, and why Jesse is cautious about connecting his own AI tools directly to ad platforms.

The real story behind Tire Agent’s compliance tool: a person spent four hours a day manually checking ad campaigns for trademark and legal risk. The team asked if AI could replace that. The answer: mostly no. The tool runs on Python and deterministic rules for 95 to 98 percent of the work, with AI handling a small remaining slice. Same employee still does the job, now in 30 minutes instead of four hours, freed up to help grow the business instead of just protecting it.

Jesse’s three-part approach to internal AI adoption: education, picking one platform and sticking with it, and using AI to build tools for the team. His magic wand wish: a self-learning system that generates ad creative at the volume the platforms now demand. And what he actually uses AI for outside work: putting his friends into funny AI-generated scenarios for a group chat.

Full episode of The New Builder.

Chapters:
0:00 Does the agent make the decision you told it to make?
0:27 Welcome Jesse Aloof
1:00 What Tire Agent actually is
1:50 How tire selection works online
4:40 Where AI touches the customer experience
6:00 Why tire buying is too high stakes to rush AI
9:30 Finding the AI champion inside a company
13:00 AI on Meta and Google ad platforms, whether you opt in or not
15:30 Why Jesse hesitates to connect his own AI to ad platforms
18:30 Inside Jesse’s workflow: education, one platform, internal tools
24:20 The compliance tool: four hours a day down to 30 minutes
28:30 Built without engineering, by a non-coder
31:00 The magic wand question: creative at scale
33:30 What Jesse actually does with AI outside work

****************************

Get The New Builder Newsletter: https://thenewbuilder.substack.com/

Connect on LinkedIn: https://www.linkedin.com/in/brianhecht/

Website: thenewbuilder.ai

Mark Fershteyn spent seven years building Recapped the traditional way: raise $8M, hire 20-something people, grind. It never became the success he thought it should be. Then ChatGPT came out — and for the first time, a non-technical founder could take an idea straight out of his brain and build it himself. He was hooked. Now his career is AI, and he's building it completely differently.

What we get into:

• The AI epiphany: "I was the idea guy who always needed other people. Suddenly I didn't."
• His first builds — grading sales calls, auto-drafting follow-up emails — and the scary realization: out of the box, AI was better than 80% of the reps he'd hired over 15 years
• The honest math: 25 employees before → 5 "specialized generalists" today, same output or more
• Agent managers vs. everyone else — his two-classes-of-workers prediction (and why he no longer thinks unemployment spikes: 10-100x more companies offset the layoffs)
• The soul-searching after Recapped: reverse-engineering happiness instead of chasing venture returns
• Rework, DHH, and why a $1-2M lifestyle business is 100x easier than the same outcome through VC
• The founder capturing the value instead of the investors — the new era of entrepreneurship we both see coming
• His compound business: Mastering AI (consulting, training, custom agents — all inbound), Zealous (his "second brain" OS that does everything he doesn't want to do), three stealth B2B SaaS products, and the North Star: his own video game studio, first game published by next year
• My pushback: didn't you just back into another venture-style company? His answer on the SaaS-pocalypse, why "AI builds it just-in-time" is a terrible UX take, and what won't get swallowed by the model (very niche or very complex)
• "I'd rather be vibe coding" — 5-10 ideas spun up a week, 90% cut, building every waking hour

****************************

Get The New Builder Newsletter and Join the Community: https://thenewbuilder.substack.com/

Connect on LinkedIn: https://www.linkedin.com/in/brianhecht/

Website: thenewbuilder.ai

Shanif Dhanani is a four-time founder and data scientist who was doing AI back when it was just called machine learning. One company acquired by Twitter, another (Apteo) backed by our accelerator, which is where we met. Now he's building Nobi, a semantic search engine for e-commerce that has never lost an A/B test.

What we get into:

• AI before it was AI — forecasting and data science in the pre-LLM era, and the prediction he got completely wrong ("machines can't understand meaning, and maybe never will")
• Why he's more impressed AND less impressed than he was in 2023. Ai is great at coding, still can't run your life
• The expectation gap: he thought he'd work less. He works 12-hour days reviewing what the agents did.
• Junior workers vs. senior workers — who AI actually displaces, and the dichotomy he's watching closely
• The LinkedIn experiment: 3 posts a day for 20 months. Did it get him customers? (Honest answer: no — and why.)
• Case study: Lucchese, the 150-year-old luxury bootmaker. "Card holder" returned 2 results on keyword search — semantic search surfaced the whole catalog, won the A/B test, and drove an extra $1M in sales in 12 months.
• Semantic search explained without jargon: conceptual search vs. keyword search
• "Nobi is not a chatbot" — real queries like "what black boots should I wear to a formal wedding?"
• Compound startups: why having side projects used to make you toxic to VCs — and why he now runs 5 ML models, 6 agents, and 3 projects at once
• The automated SEO/AEO engine producing 2-3 bottom-of-funnel articles a day (that isn't slop)
• Building for fun: he used AI to mod Civilization VI into a harder game — in a few hours, knowing nothing about modding
• And the big one: agents are extremely good at tasks, really poor at jobs. Anyone claiming otherwise is selling something.

****************************

Get The New Builder Newsletter: https://thenewbuilder.substack.com/

Connect on LinkedIn: https://www.linkedin.com/in/brianhecht/

Website: thenewbuilder.ai

I sit down with Andrew Ackerman, exited founder, reformed VC, professor...and one of the most thoughtful AI users I know, for an unfiltered conversation about how he actually uses this stuff every day.

No theory. No "AI will change everything" hand-waving. Just how a non-coder runs his work, his classroom, and his AI-centric startup with Claude.

What we get into:

• The silly little problem that pulled him into Claude (he just wanted to know when podcast episodes dropped)

• How to think about Chat → Projects → Cowork → Code as one toolkit, not four

• Why he treats Claude like training a junior employee — and why that mental model unlocks everything

• Skills vs. projects vs. agents — what's actually different and when to reach for each

• "Gen AI Jujitsu" — how he turned ChatGPT into the assignment for his entrepreneurship students (and what the data is telling him)

• Building Zero RFI: how an AI-centric startup actually runs day to day

• Meeting-prep skills, company dossier automations, and the "no one cares about the bot on the call anymore" reality

• How his daughter is vibe-coding a walking-route app

Andrew's book "The Entrepreneurs Odyssey" → https://www.amazon.com/Entrepreneurs-Odyssey-Approach-Startup-Success/dp/1032883545/ref=tmm_pap_swatch_0

https://www.andrewbackerman.com

http://www.linkedin.com/in/andrewbackerman

https://x.com/andrewackerman

https://www.instagram.com/andrewbackerman/

Todd Baldwin, CEO of Crafted, breaks down how his lean startup built a data bridge between social media marketing and in-store retail sales — and how AI agents are helping them punch way above their weight.

Crafted works with brands selling into Walmart, Target, and regional grocery stores, solving one of the hardest problems in retail marketing: if you run a $100K TikTok or Instagram campaign, how do you know if it drove actual purchases in a physical store? Todd walks through how they built a receipt-based rebate tool that closes that loop — and how the data they're collecting is now fueling an AI-powered insights engine.

Todd also shares his "Maslow's Hierarchy of AI" framework — why most startups skip the foundational steps (descriptive → predictive → prescriptive → autonomous) and jump straight to agents before they've earned the right to. Plus: how Crafted is rolling out 27 internal AI agents to keep their team lean while scaling.

If you're a Builder thinking about how to use AI strategically — not just as a shiny object — this episode is for you.

****************************

Get The New Builder Newsletter: https://thenewbuilder.substack.com/

Connect on LinkedIn: https://www.linkedin.com/in/brianhecht/

Website: thenewbuilder.ai

Jeff Latz spent 10 years as a professional engineer before walking away to pursue music — and ended up building Friday, an e-commerce platform for independent musicians, almost entirely with AI agents.

In this episode, Brian Hecht sits down with Jeff to unpack how he went from professional engineer to solo founder, why he chose Claude Code + Conductor to run multiple AI agents in parallel, and how he shipped a working Bandcamp alternative in under two weeks.

They also get into the real question behind every AI-powered startup right now: is this a business, or just a really fun experiment?

What we cover:
— Why Jeff left a decade-long engineering career to go all-in on music
— How he discovered AI coding tools and built his own workflow from scratch
— What Conductor is and why it lets him run a "team" of Claude Code agents
— The prompting philosophy that actually works (hint: bullet lists beat paragraphs)
— Friday Music: what it is, why Bandcamp failed artists, and what comes next
— The bootstrapper's take on VC money and building something that lasts

************************
Connect on LinkedIn:   / brianhecht  

Follow Me on Youtube: https://www.youtube.com/@the_new_builder

Everyone says "just use AI agents." Nobody explains what that actually means — until now.

Davida Ginter is the co-founder of ELOO, an exciting startup that helps companies quickly train and onboard AI agents to perform virtually any task. In this episode, we break down what an agent actually is (vs. just prompting ChatGPT), how to onboard one like a new employee, and why setting it free on day one is the fastest way to get bad results.

We also get into the harder questions: What does it mean to build a startup when your product could be obsolete in 3 months? How do you find your moat in an AI-native world? And is the "vibe coding" era making founders skip the most important step?
If you're a solo builder, a non-technical founder, or just trying to figure out where agents actually fit into your workflow, this one's for you.

Scott Werner has spent three years building AI-native products in one of the fastest-moving corners of tech. What he learned: being early isn't the same as being right, and building for where models are going is a trap if you can't sell what exists today.

In this conversation, Scott and Brian get into the reality of building in DevTools and AI right now — the pivot from product to distribution, why an accidental newsletter became the most valuable thing Scott built, and what the VC model gets wrong about this generation of companies.

They also talk about the new shape of expertise, why a builder who started three months ago has lived through 25% of the Claude Code era, and why the window for people who just start is wider than it looks…if you stop waiting to feel ready.

Scott is the founder of Sublayer and one of the most clear-eyed builders Brian knows.

Guest: Scott Werner, Founder & CEO of Sublayer

Topics covered:

• Why "build for where models will be in 6 months" is advice that can quietly kill your company
• The distribution problem: what Scott figured out after watching things he built go nowhere, then watching near-identical products blow up
• How an accidental newsletter and a "somebody should do this event" became his most defensible assets
• The three moats that actually matter in an AI world: deep domain expertise, proprietary data, and distribution
• Why the VC model may be mismatched to this generation of companies — and what the alternative looks like
• Audience-first business building: what it means when the product can change every quarter but the trust doesn't
• What Scott is building next (an agentic to-do manager and a place to share Claude artifacts)
• Why "you can't learn karate from a book" now applies to your AI instructor too