people overcomplicate gtm engineering here's the whole stack a postgres database with 3 tables: accounts, people, signals waterfall enrichment. apollo first, then getleads, then leadmagic, then guess first.last@ and first@ and run it through million verifier. you only pay the next provider when the cheaper one misses re-verify everything. i had an enrichment provider mark 33 emails VALID and million verifier said only 10 were actually good then an agent that runs every morning on exa looking for buying signals. new funding, hiring for the role you sell to, new tech on their site, the founder posting about the exact problem you solve it writes every signal into postgres with the url as proof score the account, push the good ones into instantly thats a gtm engineer you dont need 40 clay tables and a $2k/mo stack for this you need claude code and a weekend get started for free - graphed.com/platform?utm_sou…

the two best times to cold email someone 1. they just started a new job 2. their company just posted a job a new vp of marketing reviews every tool in the stack in their first 90 days. they want to change something fast a company hiring a revops analyst already has the problem. they just told you in public, and the job post tells you which tools they use how i'd run it with claude code pull job changes and new job posts from coresignal every day into postgres match them to your TAM so you only keep accounts that fit your ICP find the decision maker with apollo and verify every email with millionverifier push them into instantly with {{newTitle}} and {{jobPostTitle}} as variables keep the email under 75 words "saw you just started as vp of marketing at acme" beats any first line an AI writes from their linkedin the timing is the personalization get started for free - graphed.com/platform?utm_sou…

every company is going to have a cast of ai characters in their brand style guide logo, colors, fonts, and now the characters selling to creatives? your character does motion graphics for tv commercials selling to service business owners? beard, trucker hat, graphic tee, sitting in the workshop multiple characters, they talk to each other in the same ad every ad generated with seedance trademark them like a mascot or a logo mark this seems so obvious to me get started for free - graphed.com/platform?utm_sou…

what is an ai agent for marketing it's just code with a thinking loop on a live data stream. it makes decisions on top of that people think you need to put god in a box to run a facebook ads account this is absolutely the wrong way to do this i don't want an agent to think. i want it to run a system and the best way to have it run the system is discrete code then it goes and rewrites the algorithm of that code to run the account this is exactly what the human media buyer was doing 1. researching new creative 2. uploading new creative 3. turning losers off 4. promoting winners 5. making more ads like the winning content the agent should just be code that's running this system. not something that's going rogue you wouldn't want an employee to go rogue. why would you want your agent to? get started for free - graphed.com/platform?utm_sou…

marketing engineering today how to make very good avatar UGC with seed dance research the pain points and desired outcomes for your target customer by using the exa ai api to scrape reddit and other social then use these to write scripts. pro tip for the prompt is have it be hook, pain point / desired outcome driven, and a conversational tone then take this script and use something like 11 labs to do a script read as an mp3 you want to make the script read first so seed dance is only doing lip sync, not the audio render as well it reduces seed dance's mistakes on audio reads then send the mp3 with a reference image of your avatar to seed dance and this is how you can make great quality you can do this whole workflow on graphed, the marketing engineering platform install the graphed cli and just give it this post sign up free - graphed.com/platform?utm_sou…

i get asked almost every day at this point "how do i make sure my ai ads are on brand" the easiest way is a vision loop take the brand style guide make v1 of the ad with chatgpt image 2.5 have a vision model look at v1 and check it against every rule in the style guide if it fails, send it back to chatgpt image 2.5 with the vision model's notes repeat as many times as it takes until it passes the gate this is how we make ai ads work in regulated industries it's basically going back and forth with a graphic designer and giving feedback over and over, except each round takes a few seconds instead of a few days get started for free - graphed.com/platform?utm_sou…