1/ I'm an auto marketer who taught myself to code. Then I built an AI agent that cut our competitive-intel response time from 72h to 12h — and content cost by 20%+. Here's the exact stack and how it works 🧵
2/ The problem first. In an EV price war, a rival can launch a campaign at 9am and own the story by noon. Our old flow: analysts manually tracking 6 platforms, reporting the next day. By then the narrative was already lost.
3/ So I built an agent loop with Python + Playwright + MCP: - Pull our + competitors' posts, indices and sentiment in real time - Auto-tag by topic and competitor - Push a clean digest to our dashboard No human in the monitoring loop until a decision is actually needed.
4/ The real unlock wasn't scraping. It was framing. I made the agent output marketing decisions, not data: 'A rival is winning the safety narrative — here are 3 counter-angles.' Data nobody acts on is just noise.
5/ Results after rollout: - Response cycle 72h to 12h - Event content cost down 20%+ - The team stopped firefighting and started pre-empting
6/ The lesson for marketers in 2026: You don't need to become an engineer. You need to understand your workflows well enough to automate the boring 80%.
7/ I'm documenting how I use AI agents in real marketing work as I build this account.
Follow @BigYvan for the practitioner version, not the hype. What's the most painful manual workflow on your team right now?
The humanoid market goes from ~$3B (2025) to ~$28B (2030).
BYD already owns the dealer network to distribute + service them. Same playbook Tesla runs from the factory side with Optimus.
Whoever owns the channel owns the rollout.
Would a robot help you buy a car — or kill the vibe?
BYD wants 2–3 humanoid robots in every showroom within 1–2 years.
Not a gimmick. It's a go-to-market move — and most auto marketers are sleeping on it.
Here's the read 🧵
The robot isn't there to replace salespeople.
It's there to do the boring, repeatable GTM work: explain trims, walk specs, demo features, qualify intent — at scale, 24/7, in every store at once.
That's embodied AI as a sales channel.
Everyone still calls AI a "feature" in cars.
250+ engineers are in San Francisco this week arguing the opposite: AI is the car now — perception, decisioning, the whole Level 4/5 stack.
The brands still selling "horsepower + screens" are about to get lapped by the ones selling autonomy people actually trust.
What's the first EV brand that markets trust in the AI, not the spec sheet?
OpenAI just previewed GPT-5.6. It beats Claude Mythos 5 on coding benchmarks.
But the benchmark isn't the story. The price war is.
OpenAI is undercutting Anthropic right as the model gap shrinks to almost nothing.
When two frontier models are this close, the winner isn't the smartest one. It's the one that's cheapest to put inside a workflow.
Capability is table stakes now. Distribution is the moat
@miltonheyan Exactly. Silent success is the real failure mode
Fix: freshness + format checks that throw a visible "degraded" flag instead of pretending it's fine. A late signal beats a confident wrong one
1/ I'm an auto marketer who taught myself to code. Then I built an AI agent that cut our competitive-intel response time from 72h to 12h — and content cost by 20%+. Here's the exact stack and how it works 🧵
2/ The problem first. In an EV price war, a rival can launch a campaign at 9am and own the story by noon. Our old flow: analysts manually tracking 6 platforms, reporting the next day. By then the narrative was already lost.
3/ So I built an agent loop with Python + Playwright + MCP: - Pull our + competitors' posts, indices and sentiment in real time - Auto-tag by topic and competitor - Push a clean digest to our dashboard No human in the monitoring loop until a decision is actually needed.
4/ The real unlock wasn't scraping. It was framing. I made the agent output marketing decisions, not data: 'A rival is winning the safety narrative — here are 3 counter-angles.' Data nobody acts on is just noise.
5/ Results after rollout: - Response cycle 72h to 12h - Event content cost down 20%+ - The team stopped firefighting and started pre-empting
6/ The lesson for marketers in 2026: You don't need to become an engineer. You need to understand your workflows well enough to automate the boring 80%.
7/ I'm documenting how I use AI agents in real marketing work as I build this account.
Follow @BigYvan for the practitioner version, not the hype. What's the most painful manual workflow on your team right now?
@sdotboro Hey 👋 marketing lead who learned to code.
I build AI agents that do real marketing work — cut our competitive-intel response from 72h to 12h, content cost down 20%+.
Now sharing how AI reshapes auto/consumer GTM, in public. What are you building?
@kushmergedeck Hey 👋 marketing lead who learned to code.
I build AI agents that do real marketing work — cut our competitive-intel response from 72h to 12h, content cost down 20%+.
Now sharing how AI reshapes auto/consumer GTM, in public. What are you building?
@BraedendotTECH Builder here 🙌 car marketer who codes. Building AI agents for GTM + writing about AI × EV in the open. Followed — would love to see what you're working on