Ron Dahan

Co Founder CEO @ - The world's advertising, as data your AI can read.

[Ron Dahan @thRoncho]

@ZainAkrams once discovery+deal is a tool call the bottleneck moves to writing the approval policy

[Ron Dahan @thRoncho]

@KizilCo_news ads agent + an ads mcp on the same stack is the tell. tv creatives becoming tool-callable is gonna force who owns the ad object

[Ron Dahan @thRoncho]

@SergeStack @bartekjagla immutable audit + clear escalate-on-denial is the part most mcp demos skip

[Ron Dahan @thRoncho]

@matehunyor I have to show you upspring's MCP

[Ron Dahan @thRoncho]

@efraintorres 512 → 8ms is the boring win that matters once youre chaining a pile of tool calls

[Ron Dahan @thRoncho]

@coval__ same here. mcp is solid for reads / batch edits but ad set setup is still faster in the ui for me

[Ron Dahan @thRoncho]

@jamesdevonport the creative research tool in the official mcp is the sleeper imo. pairing that with higgsfield then just shipping experiments is basically the whole loop

[Ron Dahan @thRoncho]

@DavidPawlan @Muse Love it. not to mention that with upspring’s claude connector, you can do it directly from cowork

[Ron Dahan @thRoncho]

@andrew_jennings stack is clean, curious how you keep agents from thrashing the ai on broad scans

[Ron Dahan @thRoncho]

@EricTilbury_RTB this. dsp mcp is cool precisely cause the buyer still has to know what good looks like

[Ron Dahan @thRoncho]

@codyschneider wrapper illusion is real. agents still chew the same marketing api quota. warehouse for reads / api for writes feels like the only sane split

[Ron Dahan @thRoncho]

@timopowerz @YourGuyForAi write access with no kill switch is just expensive autopilot.

[Ron Dahan @thRoncho]

@IABTechLab weekend install for seller agents is kinda crazy. still feels like theyre missing market/creative context tho

[Ron Dahan @thRoncho]

@IABTechLab yeah the testing bit is the real hard part. demos are cheap

[Ron Dahan @thRoncho]

@AI_Acq ad library research is the underrated bit. once competitor angles sit in the same loop as production it stops being random

[Ron Dahan @thRoncho]

@IABTechLab buyer agent sdks are starting to feel real. still think teams trip on context + approvals before the model even matters

[Ron Dahan @thRoncho]

@Nate_Esparza @Engineering wild how fast the platform mcps are shipping. curious whether agents get real creative/context depth or mostly just campaign knobs

[Ron Dahan @thRoncho]

@lifemaximised feeding competitor urls + transparency links into the project brain is such an underrated move. annoying part is keeping that ad context fresh without re-pasting every week

[Ron Dahan @thRoncho]

@williamkast_ curious how you handle competitor creative in that stack. research connectors help but the ad library side is still usually a separate tab circus for me (we pipe live ads into the llm via Upspring when we can)

[Ron Dahan @thRoncho]

@Matt_Jacobs yeah the structured inputs bit is the whole game. without a shared way to describe creatives/offers/context, agents just freestyle and buyers get random outcomes