went on greg's pod to talk about: - All marketing is now code and tokens. One agent replaces the designer, copywriter, UGC creator, editor and motion designer. - Long-form video editing crossed the line this year. Tesseract one-shot edits a 90-minute podcast: silence, takes, framing, active speaker. - Your job is conductor, not operator. Get out of the agent's way. - Think jobs-to-be-done, not employees. Turn what the top 1% media buyer does into code with a thinking loop on live data. - "God in a box" running a Facebook ads account failed at multiple companies. - An agent needs three things: data, access, tools. - Data: a pipeline (Airbyte) into a warehouse (ClickHouse). Never read from ad APIs directly. - Direct API reads overflow context, cause hallucination, and push the agent to build its own local DB full of dupes. - One warehouse joins ads to GA4, HubSpot, PostHog and payments. You see ad-level revenue and LTV, not just conversions. - Access: Marketing API tokens for Meta and Google, write-only. All reads come from the warehouse. - Agents act like a drunk uncle at the API. Thousands of requests get accounts banned, not "using AI." - Tools: kie, fal, OpenRouter, GetLeads, Apollo. Without them the agent cannot do the middle work. - Add a swipe warehouse. Thousands of examples, not a swipe file. Sublime or Cosmos to store it. - Feed the creative-director agent Ogilvy plus 100 hours of top ad podcasts. - Static ads flow: brand URL → site style → Ad Library → Reddit pains and outcomes → 5 concepts → GPT Image 2.5 → upload to Meta as a new ad set. - Brand style guide is a markdown file. Fonts and colors. Swap it for the extracted one. - Models swap in one line. GPT Image 2.5 is best seen so far, especially for targeted edits that keep the original image. - Regulated clients: a vision model gates each creative against a rules list and loops. Two or three passes. - UGC flow: script → fal voice clone → MP3 first. Never send the script straight to Seedance. - Seedance gets MP3 + reference image + script together. That walled garden fixes mumbled words and random faces. - Render at 480p, then upscale and grain with ffmpeg. 1080p native costs about $14. Full ad lands near $5, heading to $3. - Word-level transcript vs script. A wrong word gets spliced from the MP3 at its timestamp. No re-render. - Plastic look fix: start from a real creator photo and remix age, background, ethnicity. - Captions and motion graphics come from Tesseract off the same transcript. - Next build: a vision model scores the motion graphics frame by frame and gates on 8 or 9 out of 10. - AI avatars sometimes beat the human on the same script. Nobody knows why. - Andromeda matches good creative to the buyer. Your only job is creative velocity. - Clients go from 10 ads a month to cohorts of 30. - Software pattern: an ad gets 2 cheap demos, Meta finds the two buyers and stops spending. Feed it more. - Store the Seedance JSON for every ad. Join it to Meta performance. At 1,000 ads you have cellular-level data on what works. - The whole loop runs in a hosted Claude Code harness in the cloud. 100 creatives is the same process. - Human creators layer in via Tribe or Sideshift APIs. About $40 a video plus a spend bonus. The agent is the creator manager. - Motion graphics go over the human clip the same way as the avatar clip. - A human gate can sit before anything goes live. That is where taste enters the workflow. Watch now

the fastest seo win right now is the keywords you already rank 11 to 20 for page 2. google already thinks you're relevant and nobody clicks no semrush seat needed. dataforseo + search console + claude code 1. pull every keyword you rank for from the dataforseo ranked keywords api 2. keep positions 11-20 with 50+ searches a month 3. check them against search console so you only touch queries that already get impressions 4. group by page. usually one page does all the work 5. have claude fetch the live page and audit it. count the H1s, measure the meta (google cuts around 155 characters), check it uses the words people actually search 6. rewrite the title, meta and H1 around the top 3 keywords, before and after side by side 7. ship it and check positions in search console every week watched this run on a moving company homepage this week. 6 H1 tags, a 590 character meta, and the H1 said bay area when people search san francisco gets fixed in one prompt get started for free -

just made a Marketing Engineering Crash Course for you you're going to learn 1. what marketing engineering is and why it pays $150k - $300K 2. data pipeline + warehouse and why your agent shouldn’t hit the ads APIs directly 3. the whole tech stack you need including server, s3, postgres, cron, auth, git, api gateway 4. how to give your whole stack to claude code in one install 5. static ads creation with unified api to make 1080x1080 ads with chatgpt image 2.5 6. meta api token creation system user, never expires facebook ads, so you can upload ads, turn ads off one prompt across meta 7. data analytics of ga4, search console straight into a dashboard 8. linkedin engagers to emails: scrape, waterfall enrich, million verifier 9. seo strategy for saas: x vs y keywords + page 2 keywords from search console 10 . then blog posts on autopilot: research to published with to your CMS with an SEO plugin 11. making AI avatar ugc ad scripts: reddit pain points via the exa api

chatgpt is now indexing the web they're indexing the web like google researchers found its name in ChatGPT's traffic called Labrador from May 21 to July 21, every search result in ChatGPT carried a label for where it came from three of the labels were scraping services that pull Google results. the fourth was Labrador, which is OpenAI's own index a few things stood out to me it barely overlaps with Bing. Researchers ran the same searches on both, and only about 1.5% of Labrador's results were in Bing's top 20 if you've been treating your Bing rankings as your ChatGPT rankings, they're not the same thing. free users mostly get Labrador in the default Instant mode, ChatGPT didn't open a single page in 93% of answers it answered from a stored snippet of about 200 characters Paid Thinking mode is different about 75% of its results came from scraped Google, and it actually reads the pages your snippet comes from your H1 and the text right around it ChatGPT ignores your meta description, and the snippet doesn't change until the page gets re-indexed updating a page doesn't update ChatGPT. it keeps a cached copy of your page. pages nobody asks about can stay stale for months. it also doesn't run JavaScript, so anything that loads with JS isn't there OpenAI runs three separate bots. GPTBot collects training data. OAI-SearchBot decides whether you show up in ChatGPT search. ChatGPT-User fetches a page live during a chat. I keep seeing sites that blocked all three to stay out of training and then wonder why ChatGPT never cites them so what I'd do: allow OAI-SearchBot in robots.txt. make sure your H1 still makes sense if someone reads only that line answer the question in your first paragraph put the date and author in visible text near the top render your content on the server ChatGPT still mixes in Google and Bing results. But OpenAI is already testing "prefer our own index" in shopping, so I'd expect more answers to come from Labrador over time Research credit: Peec AI, Search Engine Land, RESONEO, a lot of X posts

AI search explained ranking on chatgpt vs claude vs gemini Claude = Brave Search Gemini / AI Overviews = Google ChatGPT = a hot mess of bing + its own index + google + other feeds Brave is on Anthropic's subprocessor list, and there's a BraveSearchParams field inside Claude. Studies found about 79–87% of the URLs Claude cites come from Brave's top 10 organic results, and Claude barely re-ranks them. If you rank on Brave, you get cited with that said Claude only searches on about 37% of prompts. ChatGPT searches on about 90%. Claude looks things up when a query needs fresh info, rankings, a location or a comparison. Otherwise it answers from training data gemini page one on google, your in gemini. pretty simple chatgpt, high on the pretty crazy scale It started on Bing, and some of that is still there. But OpenAI now has its own index, built by OAI-SearchBot, plus third-party scrapers and partners like Yelp and TripAdvisor. How closely it matches any one search engine's top results is low and keeps changing. When you ask Claude and ChatGPT the same prompt, only 8–20% of their citations overlap So being dominant in one engine doesn't mean you show up in the others. What to do about it: Track Brave rankings separately. If you want Claude citations, this is your most direct lever. Submit your URLs to Brave directly so they get picked up faster. For ChatGPT Don't skip Bing. It's cheap insurance for ChatGPT: set up Bing Webmaster Tools, turn on IndexNow and allow OAI-SearchBot in robots.txt. also go past rankings. Get brand mentions, write structured and authoritative pages, and get lots of other sources saying the same things about you. Write for the prompts that trigger search. Claude mostly searches on comparisons, "best X," recent news and local queries. Build pages for those. Remember when people optimized for Google and Bing separately? That's basically back

new podcast episode out and you'll learn... how the most bloated software with extensive API endpoints is now arbitrage spent ten years hearing that salesforce is bloated enterprise garbage and you should buy something clean instead. attio, hubspot, whatever shipped most recently changed my mind on this about six months ago and i've bought software differently ever since you can plug the salesforce CLI into claude code and provisioned user security settings from the terminal that sounds boring unless you're a salesforce admin what it actually means is every one of those legacy endpoints, all the metadata about what each object does, all that documentation nobody wanted to read an agent consumes all of it fine. the ugly admin surface that made salesforce miserable to click through is the surface a model needs to get work done we ran hubspot at a previous company. nice UI. you hit a wall the second you want to do something through the API that you can do in the app so now i ask one question before buying anything can i do everything through the API that i can do in the UI if the answer is no it's off the list, doesn't matter how good the product looks and if i need an interface i'll build one. cost of code is basically zero, my team gets the exact view they need when they need it listen to the full episode at the link

i can't express to you how stupidly powerful claude code is for SEO with an .env file containing your keywords everywhere API key and your dataforseo API key some things you can do for your saas company pull your full keyword universe using keywords everywhere's related keywords and people also search for endpoints then send that entire list to dataforseo's SERP API to see who's actually ranking and where the gaps are. full content calendar with clustering and prioritization generate programmatic landing pages at scale. if you're a saas serving 20 industries you hit keywords everywhere for the long tail variations per vertical then check dataforseo for difficulty and SERP features on each one and claude code just generates unique pages with the right semantic terms and schema markup already baked in link building using dataforseo's domain intersection endpoint that shows you every site linking to your competitors but not to you. pull backlink profiles for your top 5 competitors run the intersection find the gap scrape contact info and draft personalized outreach emails referencing the specific page they link to. entire pipeline in 8 minutes build internal linking maps using keywords everywhere's related keyword data to create topical relevance clusters then have claude code generate the actual linking structure across your site. not random links. real semantic relationships that google rewards run a full technical audit using dataforseo's on-page API and have claude code automatically generate the fix for every issue it finds. missing canonicals broken schema thin content orphan pages. it finds the problem and writes the code to fix it if you want this get it below

just pick a business vertical tattoo removal permanent makeup nutrition coaching lactation consulting med spas build vertical agent software for them that manages their website handles form submissions does chat support automatically grows their google maps listing impressions has scheduling handles phone calls then have agents run your gooogle ads facebook ads cold email SEO AI search social media management and you have a $80k MRR business

This is actually the way. A year ago we started building the GTM super app at Rippling…today we have over 2,500 internal users across sales, marketing, FDE, analytics, finance, ops powering more use cases than we could have imagined. Our leverage wasn’t necessarily the app itself, but the mono repo that managed applications, data pipelines, jobs, storage, evals, ML models in one place. We can build new agents through Codex sessions with CLI tools to bring in external context from our warehouse, jira, docs and ship it in less than a week.

last week went on gregs pod you probably missed it bc you're super busy here's everything you'll learn LinkedIn engagement is a hand-raise signal, beats firmographics Track 10–20 category outliers, get 80% industry coverage Use your for-you feed, not search, sourcing creators Business accounts work as lead sources, not just people Apify API Maestro actors: most stable LinkedIn scrapers Post reactions plus post comments = full engager list Daily cron pulls net-new posts, then extracts engagers ICP-fit research happens before enrichment, not after Waterfall enrichment: cheapest accurate source first, expensive last Origami aggregates the entire waterfall behind one call LeadMagic specifically for mobile phone number enrichment MillionVerifier before sending, or deliverability dies fast Buying broker contact data is legal; usage rules differ Burner domains protect your core domain deliverability Four domain buckets: cold, marketing, transactional, business Roughly $200/month total to start sending 10k Instantly webhooks push positive replies to your agent Inbox agent answers questions, drives toward booked demos Program six-month re-touches to revive gone-cold leads Wire calendar access so agent verifies actual bookings Agent equals code, thinking loop, live data stream Don't pay tokens for what cheap CPU does Agent frameworks are usually bloat for finite problems Model the human's real process, not autonomous god-box Organic: interviews, sales calls, Slack, Notion, Gong "Write good LinkedIn content" produces mid, flaggable slop Best content is already trapped in internal conversations Ordinal schedules across accounts, feeds analytics back in Analytics stream tells agent what to snowball or remix Repost proven winners every 90 days, never daily Find what market wants, then build; same for content LinkedIn paid impressions run ~$22 per thousand — earned media Topic-based pages work when personal brand isn't appealing Timestamps 2:28 — Agent one: signal-based cold outbound 9:11 — Apify scraping and extracting engagers 15:47 — Waterfall enrichment, validation, compliance 21:40 — Inbox infrastructure, sending, inbox-managing agent 31:41 — Agent two: organic LinkedIn content engine