AI DTC WTF

Question: How do you use Claude and Codex without getting locked into one platform? Answer: Keep your skills and instructions outside any single tool and point every tool at the same copy. Dustin Thompson stores his skills on GitHub, symlinks his CLAUDE.md file to his AGENTS.md file so Claude Code and Codex read identical instructions, and uses GitHub versioning to roll back a broken skill in plain English. Once the skills live in one canonical place, choosing Claude or Codex for a project, or moving a scheduled task between them, becomes a feature and usage decision rather than a migration. In this episode, Andrew Foxwell and Thomas Moen talk with Dustin Thompson, owner of Konnected Interactive, about how to use Claude and Codex side by side without getting locked into either platform. Dustin explains why every skill he has lives in one place on GitHub, how a symlink lets Claude Code and Codex read the same instructions, where scheduled tasks should run, and how he tests a new model before trusting it with real work. Dustin owns Konnected Interactive, where he builds AI systems for agencies and ecommerce teams, and he is the builder behind the Foxwell Bot and much of the internal tooling at Foxwell Founders and Foxwell Digital. We discuss: - Why Dustin runs Claude and Codex all day instead of picking one - How Codex imports your memories and skills from Claude - Why your AI skills belong on GitHub instead of Dropbox or a local drive - Using GitHub versioning to roll back a broken skill in plain English - Sharing one skills library across a team, and seeing which skills actually get used - What a symlink is, and how CLAUDE.md and AGENTS.md stay in sync - Whether scheduled tasks should run in Claude or Codex - Why reporting is the first thing an agency should automate, not ad creative - Why AI without a strategy usually means more cost for the same output - How to test a new model with known-answer tasks, reasoning traces and evals Listen to the full podcast: https://podcasts.apple.com/us/podcast/ai-dtc-wtf/id1896687855 Show site: https://www.ai-dtc-wtf.com/ Connect the Foxwell MCP to Claude or ChatGPT: https://www.foxwelldigital.com/mcp Guest: Dustin Thompson, Owner, Konnected Interactive Konnected Interactive: https://konnectedinteractive.com/ Dustin on LinkedIn: https://www.linkedin.com/in/dustinjthompson This Episode is sponsored by Motion's very own Runneth. Runneth is the AI Tool that is quickly becoming the go to decision maker brands turn to when it comes to creative decisions within their ad accounts. From Creative strategist to hook inspirations and creative briefs, runneth can do it all. See the case studies and try runneth yourself https://runneth.com/ AI DTC WTF is the show for brand owners, in-house advertisers and agency operators who use AI every day and want tactical workflows they can implement now. Published by Foxwell Digital: https://www.foxwelldigital.com/ To connect with Andrew Foxwell reach him here [email protected] To connect with Will Sartorius DM him here https://x.com/will_sartorius To connect with Thomas Moen DM him here https://x.com/thomasmoen To learn more about Foxwell Founders and conversations like this one, go here: https://www.foxwellfounders.com Learn more about Foxwell Digital and join our newsletter: https://www.foxwelldigital.com/email Chapters 02:22 Why run Claude and Codex side by side 05:52 Why your AI skills belong on GitHub, not Dropbox 08:03 How to share one skills library across a team 10:49 What a symlink is, and why it keeps Claude and Codex in sync 11:53 Should scheduled tasks run in Claude or Codex? 14:16 What to automate first: reporting, not ad creative 18:55 Why AI without a strategy just costs you more 21:28 How to test a new AI model before you switch 24:35 What Dustin is building: the Foxwell Bot and a model orchestrator

Is AI making you a worse marketer? AI makes you a worse marketer when you use it to skip work you never learned to do, and a better one when you use it to speed up work you already understand. Ben Feys runs PrettyBoy with one co-founder and holds every AI build to three tests: the result has to be measurable, it has to beat the alternative, and it has to save money. He automates research, planning and briefing on workflows he knows end to end, keeps people on the creative judgment, and tracks AI-made ads separately from human-made ads so the performance question gets settled by data. In this episode, Andrew Foxwell and Will Sartorius ask Ben Feys, co-founder of PrettyBoy, how a two-person DTC brand uses AI to get more done without losing the skills that make the work good. Ben explains the three tests every AI build has to pass, why he tracks AI-made ads separately from human-made ads, and why he thinks human-written content could be a small brand's moat. PrettyBoy is a men's skincare brand run by Ben and his co-founder Kevin Niehoff. Ben automated the brand's email workflow end to end with Claude, built a company brain on top of their ad, retention and creative data, and has strong opinions about where that stops paying off. We discuss: The three tests a two-person team uses to decide whether an AI build is worth it Automating an email workflow with Claude, from performance analysis to briefs loaded into ClickUp Why vibe coding your own version of a SaaS tool is misguided for most operators How automating creative entirely makes you settle for "nice enough" ads Tracking AI-workflow ads separately from human-workflow ads to see which actually perform What a company brain is good for, and the two kinds of people who benefit most Why AI works best for people who already know how to do the job Why hiring for "Claude native" backfired, and what to hire for instead How to use AI as a creative strategy teacher instead of a shortcut Why going all in on human-made content could be a small brand's moat Show site: https://www.ai-dtc-wtf.com/ Connect the Foxwell MCP to Claude or ChatGPT: https://www.foxwelldigital.com/mcp Guest: Ben Feys, Co-Founder, PrettyBoy PrettyBoy: https://www.yoprettyboy.com/ Ben on X: https://x.com/fenbeys Ben on LinkedIn: https://www.linkedin.com/in/benjamin-feys-23112950/ This Episode is sponsored by Motion's very own Runneth. Runneth is the AI Tool that is quickly becoming the go to decision maker brands turn to when it comes to creative decisions within their ad accounts. From Creative strategist to hook inspirations and creative briefs, runneth can do it all. See the case studies and try runneth yourself https://runneth.com/ AI DTC WTF is the show for brand owners, in-house advertisers and agency operators who use AI every day and want tactical workflows they can implement now. Published by Foxwell Digital: https://www.foxwelldigital.com/ To connect with Andrew Foxwell reach him here [email protected] To connect with Will Sartorius DM him here https://x.com/will_sartorius To connect with Thomas Moen DM him here https://x.com/thomasmoen To learn more about Foxwell Founders and conversations like this one, go here: https://www.foxwellfounders.com Learn more about Foxwell Digital and join our newsletter: https://www.foxwelldigital.com/email Chapters 00:00 Is AI making you a worse marketer? 02:28 Automating an email workflow with Claude, from analysis to ClickUp briefs 05:16 Why vibe coding your own SaaS stack rarely pays off 07:00 Does automating ad creative make your ads worse? 08:55 What a company brain is good for, and who it helps most 15:29 Why AI works best when you already know how to do the job 19:54 Why hiring "Claude native" strategists backfired 23:25 What to look for in a creative strategist now 29:29 What running a DTC brand looks like a year from now 32:54 Why human-made content could be a small brand's moat (02:28) - Automating an email workflow with Claude, from analysis to ClickUp briefs (05:16) - Why vibe coding your own SaaS stack rarely pays off (07:00) - Does automating ad creative make your ads worse? (08:55) - What a company brain is good for, and who it helps most (15:29) - Why AI works best when you already know how to do the job (19:54) - Why hiring "Claude native" strategists backfired (23:25) - What to look for in a creative strategist now (29:29) - What running a DTC brand looks like a year from now (32:54) - Why human-made content could be a small brand's moat

How do you hire creative strategists and copywriters when AI writes every work sample? Stop judging creative strategists on work samples, because you can no longer tell what the candidate wrote and what AI wrote. Make them do the actual job instead: FireTeam's AI interviewer hands candidates a deliberately flawed brief, asks them to find and fix the problems, pushes on why each fix is right, and flags pasted answers. Past that screen, the strongest signal is proof that someone builds and tests with AI while staying skeptical of its output, such as a steady record of GitHub commits. In this episode, Andrew Foxwell and Will Sartorius talk with Jess Bachman, co-founder of FireTeam, about how to hire creative strategists and copywriters when AI writes nearly every work sample. You'll learn how Jess's AI interviewer tests how candidates think, how FireTeam's AI brief writer outperforms a strong human copywriter, and why he would judge a creative strategist by their GitHub commits. Jess runs FireTeam, a performance creative agency that gets paid when ads win. He deployed the AI interviewer to hire a copy engineer and has iterated on it since, and every FireTeam brief now runs through the agency's AI brief writer. We discuss: Why work samples stopped being a useful way to hire creative talent How an AI interviewer uses a deliberately broken brief to show how a candidate thinks Paste detection, dictation false positives, and whether AI screeners turn off good candidates How FireTeam proved its AI brief writer beats a human copywriter, using its own win data Why AI is still terrible at coming up with ad ideas, and what the team does instead Why client learnings, not the prompt, drive 75% of the results Why Jess keeps performance analysis entirely human Why the creative volume game is not coming from Andromeda Why AI raises the floor to a C but cannot turn someone into a builder The hiring signal Jess trusts most for AI-era creative strategists Get the newsletter: https://www.foxwelldigital.com/email Listen to the full podcast: Apply to Foxwell Founders: https://www.foxwelldigital.com/apply Guest: Jess Bachman, Co-Founder, FireTeam FireTeam: https://fireteam.is Jess on X: https://x.com/HireFireTeam Jess on LinkedIn: https://www.linkedin.com/in/jessbachman/ To connect with Andrew Foxwell reach him here [email protected] To connect with Will Sartorious DM him here https://x.com/will_sartorius To connect with Thomas Moen DM him here https://x.com/thomasmoen To learn more about Foxwell Founders and conversations like this one, go here: www.foxwellfounders.com Learn more about Foxwell Digital and join our newsletter: https://www.foxwelldigital.com/email This Episode is sponsored by Motion's very own Runneth. Runneth is the AI Tool that is quickly becoming the go to decision maker brands turn to when it comes to creative decisions within their ad accounts. From Creative stratigest to hook insporations and creative briefs, runneth can do it all. See the case studies and try runneth yourself https://runneth.com/ Chapters 00:00 How to hire creative strategists when AI writes every sample 02:17 Inside FireTeam's AI interviewer and its deliberately broken brief 04:16 Paste detection, and whether AI screeners scare off good candidates 07:21 How an AI brief writer beat a human copywriter 10:44 From a brainstorm gist to a finished brief in 30 minutes 16:33 Why your data, not your prompt, makes AI copy win 21:29 Does AI only help people who could already write? 22:24 Why the creative volume game was never real 27:50 Can AI turn an average employee into a 100x employee? 31:52 Why GitHub commits are the new creative strategist portfolio

How do you build a business operating system with Claude Code? You build it one component at a time and connect them, rather than building isolated dashboards. Start with the smallest painful process you already pay software for, rebuild it so it writes to your own database, then make the next tool read from that same data. Joe Chura rebuilt his brewery's sensory testing app in a single evening in Claude Code, tied it to batch numbers, then layered on customer support, wholesale onboarding, creative production and retail campaigns until every system in the company shared one set of records. The test for each piece is whether it increases revenue or reduces cost, not whether it is impressive. In this episode, Andrew Foxwell and Will Sartorius talk to Joe Chura about how to build a business operating system with Claude Code instead of buying and stitching together software. Joe rebuilt his brewery's production, support, wholesale, creative and retail systems himself in about 90 days, and explains how the pieces connect, which builds paid off fastest, and why isolated dashboards are the mistake most operators are making right now. Joe founded a digital marketing agency in 2011, grew a software company from $0 to $40 million in revenue in five years, sold it to Cars.com, and ran a 1,000-person organization with 100 developers. He now runs Go Brewing, a non-alcoholic brewery he says was the fastest growing in the country last year, with three salespeople. We discuss: - Why Joe rebuilt a cheap sensory testing app instead of renewing it, and what that unlocked - How to connect internal tools so batch data, customer complaints and cost anomalies all read from one source - Turning customers into a sales force with an ambassador app that pays out on validated leads - Why the rate-of-sale math on non-alcoholic beer makes traditional field reps uneconomical - A full screen-share walkthrough of Brewery OS, from nightly bug sweeps to a tank planner that saves a week of work - The three highest-ROI things to build with Claude Code, starting with your Shopify theme - How connecting GitHub to a Shopify theme took Go Brewing's conversion rate from under 5% to 10% in a weekend - Using Claude Design to build a design system, then handing it to Claude Code as a prompt - Why reading Microsoft Clarity every day is the cheapest optimization loop in ecommerce - The mindset shift that is harder than the technology: getting a team to stop asking for reports Never Miss and Episode: https://www.ai-dtc-wtf.com/aidtcwtf-newsletter To connect with Joe Chura, Founder and CEO, Go Brewing Go Brewing: https://gobrewing.com/ Joe on LinkedIn: https://www.linkedin.com/in/joechura/ To connect with Andrew Foxwell reach him here [email protected] To connect with Will Sartorious DM him here https://x.com/will_sartorius To connect with Thomas Moen DM him here https://x.com/thomasmoen To learn more about Foxwell Founders and conversations like this one, go here: www.foxwellfounders.com Learn more about Foxwell Digital and join our newsletter: https://www.foxwelldigital.com/email This Episode is sponsored by Motion's very own Runneth. Runneth is the AI Tool that is quickly becoming the go to decision maker brands turn to when it comes to creative decisions within their ad accounts. From Creative stratigest to hook insporations and creative briefs, runneth can do it all. See the case studies and try runneth yourself https://runneth.com/ AI DTC WTF is the show for brand owners, in-house advertisers and agency operators who use AI every day and want tactical workflows they can implement now. Published by Foxwell Digital: https://www.foxwelldigital.com/ Chapters 04:19 What was the bottleneck before Claude Code? 06:55 Rebuilding the first internal app in Claude Code in one evening 09:50 Turning customers into an extension of the sales force 14:39 Sponsor: Runneth from Motion 15:50 How to decide what to build next, and in what order 19:02 Why team mindset is the real bottleneck, not the technology 24:21 Screen share: a full tour of Brewery OS 35:53 Live demo: building a retail campaign end to end 39:40 The three highest-ROI things to build with Claude Code

Is Fable 5.1 actually better, and what should a marketer use it for? Fable 5.1 is a net improvement on Fable 5 for marketers, mainly because it is more concise and burns fewer tokens, which makes it cheaper to run for the same work. Its highest-value use is not writing copy. It is auditing the skills, agents, and markdown files you already built, finding the redundancies that were making them skip steps, and then running the cleaned-up prompts on a cheaper, faster model. Video work and translation are dramatically better than they were a few months ago. Image generation is not, and dedicated tools still win there. In this episode, Andrew Foxwell, Thomas Moen, and Will Sartorius break down what actually changed in Fable 5.1 and what it is genuinely good for. You'll learn why it is cheaper to run than Fable 5, why the best use is auditing the skills you already built rather than writing new copy, how to route tasks across different models instead of paying for the expensive one every time, and why video translation went from unusable to production ready. We discuss: - What changed between Fable 5 and Fable 5.1, and why it costs less to run - Why the highest-value use is auditing your existing skills, agents, and markdown files - The prompt audit that found the redundancies making skills skip steps - Why you should run cleaned-up prompts on a cheaper model instead of the flagship - How to route specific tasks to specific models: planning, coding, research, execution - Why less instruction now produces better output, and the Meta ads parallel that explains it - The claim that Fable 5.1 copy carries a detectable watermark, and what that means for blogs and newsletters - Why the writing finally stopped reaching for bold headers and three-item lists - The rise of the AI-forward individual contributor, and how to become one - Why one agency won't let a creative strategist touch an AI tool until they can do it analog - Why Thomas's agency stopped hiring junior people entirely - Localizing an entire instructional video library by dropping MP4 files in - Why image generation still belongs in HiggsField and not here - Tagging a folder of B-roll with Gemini Flash and cutting mashups in Claude Code To connect with Andrew Foxwell reach him here [email protected] To connect with Will Sartorious DM him here https://x.com/will_sartorius To connect with Thomas Moen DM him here https://x.com/thomasmoen To learn more about Foxwell Founders and conversations like this one, go here: www.foxwellfounders.com Learn more about Foxwell Digital: https://www.foxwelldigital.com All episodes: https://www.ai-dtc-wtf.com Join The AI DTC WTF Newsletter: https://www.ai-dtc-wtf.com/aidtcwtf-newsletter Join the Foxwell Digital email list: https://www.foxwelldigital.com/email This episode is brought to you by Runneth by Motion: https://runneth.com/ AI DTC WTF is the podcast for people who work in DTC, use AI, and want to know what is actually working right now. Every week: tactical tips, strategies, and ideas you can implement in your AI workflows to make your brand or agency more money. Chapters 00:53 Why Fable 5 was too greedy with tokens 01:30 Fable 5.1 is more concise and cheaper to run 01:45 Which model to use for planning versus doing the work 02:12 Why Anthropic keeps botching the marketing 02:26 The limits announcement that said 50% and meant 17% 02:42 Why skills kept skipping steps 03:23 The prompt audit that found the redundancies 03:53 You don't need the expensive model for everything 04:42 How to stay agnostic and route tasks across platforms 05:20 What Grok is genuinely faster at 06:20 Using Grok for small coding and research tasks 06:43 How to actually tell a model which model to use 07:14 Why stripping down your markdown files gets better output 07:31 The Meta ads parallel: less micromanaging, better results 08:19 Auditing a brief flow that was writing copy before finding personas 09:01 The watermark on Fable 5.1 copy, and what it means for blog posts 09:19 Why the writing finally stopped sounding like AI 10:53 The rise of the AI-forward individual contributor 11:47 Why creative strategists here can't touch the tools yet 12:18 AI as a compass, not a destination 12:39 How to learn creative strategy the old fashioned way 13:41 Amplifying good people versus replacing them 14:54 Why this agency stopped hiring junior people 15:49 Why Europe is further behind than the feed makes it look 16:56 Video and image generation on Fable 5.1 17:41 Localizing a whole Norwegian video library in a week 18:15 Why better reasoning makes project orchestration better 18:56 Dropping MP4 files in and getting a finished English version 20:03 Why images still belong in HiggsField 20:51 Tagging B-roll with Gemini Flash and cutting mashups in Claude Code