
who called it retention analysis and not “churn down for what”
✓GTM Engineer @modal, prev employee #1 @pavecomp

who called it retention analysis and not “churn down for what”

my philosophies for building modals GTM stack: Code-first: the best substrate for business logic is code, and the code writes, reviews, deploys, tests, maintains itself. the question isn't "can claude code use it" - it's "can i manage this solution within our SDLC". The only way you benefit from the shift from coding agents to software factories is if you treat your gtm systems as software Separation of concerns: find the best independent vendors for Data, Orchestration, and Agents. bundling is a trap - its impossible for a single vendor to be the best in every category and the bundled solution has 0 incentive to be interoperable with better vendors, so you're just trapped in mediocrity as your switching costs steadily increase over time following these philosophies led us to the following wonderful solutions: Data: @DeeplineCLI (people enrichment), Sumble (technographic info), @p0 (firmographic info) Orchestration: Modal, dbt, snowflake. One destination and SOT for GTM and product data. Agents: @hone. self-learning, highly autonomous agent platform. Uses data + tools above to automate work


they say a man never crosses the same bridge twice, for he is not the same man but after 3 times, surely the man has demonstrated some consistency? and at 12 times, you seriously expect me to believe that that’s a different guy than the 11th time? yeah right. we’ll see about that one buddy


I was a Republican city councilman. I flipped burgers at 14 to pay my way to the RNC. I named my dog Mitt. To my Republican family: you don't have to become a Democrat to vote against Ken Paxton.

@bot figure it out make no mistakes

dj bars (karaoke bars but with decks) would revolutionize the world for people in manhattan who have their usb on them at all times and have a bac of .08% and .12%

as a pathological people pleaser, i feel so much happier now that i can just say yes to everything and have my agent do it


you too can save $200k/year on conversation intelligence tools by creating a @modal account, pasting this thread into codex, and getting started with these examples: modal.com/docs/examples/batc… modal.com/blog/low-latency-v…

it took about ~6hr and ~$133 in tokens, but its not vaporware - this pipeline is live and processes new recordings within 5min of hitting the GDrive

@modal has every primitive you need to build and scale a conversation intelligence pipeline like this for $0.10/call - Google Meets recordings are loaded into a Modal Volume from GDrive - recordings are processed using open-weight diarization (@pyannoteAI) and transcription (@OpenAI) models running on Modal L4/A100 GPUs - transcripts are tagged using a Modal Shared Endpoint for KimiK3 - results stored in @Snowflake

gong (like most GTM intelligence apps) is effectively a data pipeline with a front end - calls get recorded via a bot and saved to storage - recordings are processed with a diarization (speaker partitioning) and transcription pipeline - transcript segments are tagged using LLM prompts - the LLM output is used to update the CRM

