Riley Brown walks his cofounder through nine Codex skills that turn a coding agent into a marketing department — scraping competitors, cloning their ad creative, and drafting outreach at scale.
A coding agent becomes a marketing department once it's wired into reusable skills that chain external APIs together, so it scrapes proven examples first and only then is asked to write, design, or schedule anything.
Riley Brown and his cofounder Vishal walk through nine Codex skills that turn a coding agent into a marketing operation. Skills chain APIs together: ScrapeCreators pulls competitor social content and engagement data into Notion, Foreplay finds the longest-running (proven) ads, Paper rebuilds winning ad creative with a new brand's scraped assets, and Buffer, Cal.com, and Gmail handle scheduling and outreach drafts. The throughline: AI can't originate good content, but it's excellent at copying patterns from real examples, so every skill's job is feeding the agent proven material before asking it to create anything. The result is a repeatable pipeline from research to publish-ready draft, run almost entirely inside chat.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →
Riley introduces the premise — he runs his startup inside Codex using GPT-5.6 — and his cofounder Vishal, who's shifting from engineering into marketing now that Codex handles more of the coding.
Riley fires off a ScrapeCreators skill that pulls a target creator's top Instagram videos, transcripts, and raw files into a new Notion database, then explains why AI needs real examples to write good content.
Riley shows the Chorus platform and how to add his pre-built agent (with all nine skills) to iMessage, including switching models by texting 'switch to Fable'.
The scraped creator's best-performing (non-sponsored) videos are turned into a named skill — 'Callaway Top Performing' — so any future prompt can generate new scripts written in that creator's exact voice and tone.
Using the Foreplay skill, they scrape the longest-running video and static ads from four AI competitors (Claude, Perplexity, ChatGPT, Replit) into a Notion database, with the agent analyzing why each ad likely performs well.
All the scraped static ads get pushed onto a Paper design board as a swipe file, which they then use as templates to generate new ads for their own product, Chorus.
A Firecrawl skill scrapes the Chorus landing page and pulls its logo, wordmark, and image assets directly onto the Paper board, which they then use to reskin a competitor's ad template with their own branding and copy — 'Built a $100,000 agent for $50.'
The new ad image is scheduled as a draft Instagram and LinkedIn post via a Buffer skill, with the caption style pulled from another creator's Instagram captions using the ScrapeCreators skill again; Typefully is mentioned as the equivalent for managing multiple X accounts.
An 'Agent Native Drive' skill understands the show's folder hierarchy and can create a new guest folder, hook doc, and interview questions on command, always returning direct links to whatever it created.
A Cal.com skill regenerates Riley's booking link to expose availability only within a specific future date range for a high-value podcast guest (the demo uses Shaquille O'Neal), plus drafts the outreach email with the custom link.
An email-draft skill finds the right contact for a hard-to-reach guest, drafts a Gmail message with a link he can edit and send, then — in the highest-leverage demo — batch-processes every unanswered product-pitch email from the last two months into ready-to-send decline-and-counter-pitch drafts.
Riley and Vishal recap all nine skills on screen and debate which combination is most valuable — the scraping-and-content skills (1-5) for output quality, the operational skills (6-9) for running marketing like a real job.
AI still can't originate good content on its own, so every workflow here is really the same move — scrape a real, proven example first, then hand it to the agent as the template for new work.
“The reason why AI is bad at writing content scripts is because, with coding, it's very easy to verify whether something is good or something is bad. It's way more deterministic, it's easier to validate. Whereas content, it's subjective.”
“The only thing you need to do in order to create really good content is provide really good examples.”
“If you run an ad for nine months, presumably they're spending a lot of money keeping it alive for a good reason.”
“Anything public on the Internet, Codex can scour.”
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
The full workflow list Riley walks through in order, each demonstrated live inside Codex against real accounts and real competitors.
“You can also hire the Riley Brown agent... this agent is free to try... Link is in the description.”
Soft mid-roll CTA for Chorus (his own agent platform) woven into the natural demo flow rather than a hard sales break; repeated briefly again near the end pointing to the skill downloads in the description.
00:00
00:34
01:0201:24
02:03
02:31
02:51
03:17
03:43
04:14
04:44
05:0305:40
06:02
06:39
06:5307:24
07:56
08:34
08:45
09:24
09:47
10:1010:36
11:10
11:30
12:10
12:28
12:59
13:32
14:00
14:27
14:55
15:1515:41
16:17
16:40
17:0117:31
17:59
18:36
19:08
19:28
19:48
20:25
20:39
21:10
21:37
22:0922:31
23:11
23:50
23:52
24:23
24:5725:20
25:44
26:3226:42
27:09
27:36
28:18
28:34
29:0429:42
29:54
30:35
30:53
31:13
31:49
32:10
32:40
33:09
33:31
33:5934:38
35:07
35:36
35:50
36:29
Video of the Day23:10Claude Desktop and the Codex app both quietly shipped native, multi-tab in-app browsers within days of each other — used right, one prompt now opens, drafts, and stages a dozen tabs at once.
July 15th
47:51Riley Brown and Ras Mic dig into GPT-5.6, Codex's background computer-use, and why self-scoring agent loops are turning coding tools into a general operating system.
July 12th
15:25One creator puts xAI's new Grok 4.5 through four real builds inside Cursor — and finds it roughly four times cheaper than Claude's Opus-tier pricing.
July 9th
23:58A YouTuber wires Claude Code into an agent-native design canvas called Paper and lets it generate thumbnails, Instagram graphics, and a full pitch deck from plain-English prompts.
July 6th
21:08A step-by-step build of a talking, screen-controlling desktop AI companion — no coding experience required, just long prompts into Cursor.
July 1st
39:16A 39-minute live build: one AI agent cloned from a template, one built from scratch, both running 24/7 in Slack and iMessage using the GRASP framework.
June 28th