All essays · 11
Sorted: newest first
WhatsApp bot automations with an agent dashboard (what we actually build)
The architecture behind a WhatsApp agent that qualifies leads 24/7 and the dashboard that proves it works.
Tarun ChelumallaThe Meta Sales Agent: lead form → voice agent → booked call
Our Meta Sales Agent the end-to-end automation that calls a fresh lead within minutes and books the meeting.
Tarun ChelumallaWhy every automation should ship with a dashboard
Stop hoping your workflow ran. The case for analytics built into every agent you deploy.
Tarun ChelumallaHow Claude Code can automate your codebase, without breaking it
Tactical guardrails, scoped agents, and the prompt-as-spec pattern we use to ship Claude Code workflows that don't silently rewrite production.
Tarun ChelumallaWe built a 4-system content engine. Here's the actual stack.
Not a deck. The literal tools, prompts and pipes behind 30 days of content for a D2C brand.
Vaidehi Khuntn8n in production: the 7 nodes we reach for in every pipeline
After 60+ shipped flows, these are the primitives that survive the edge cases.
Tarun ChelumallaContent at machine scale: 800 pieces from one source brief
A teardown of the system we used to ship 800 ad-ready content pieces in 30 days for a skincare brand.
Vaidehi KhuntWhy we don't offer retainers (and what we ship instead)
Retainers reward bloat. Project-based shipping rewards finished systems. Here's how we structure the work.
Vaidehi KhuntPrompt-as-spec: writing prompts the way engineers write tickets
Inputs. Constraints. Acceptance criteria. The bridge between PMs and AI agents.
Tarun ChelumallaCampaign pipelines that don't break on the third launch
Idempotency, retries, and human-in-the-loop checkpoints. The boring parts that make the magic ship.
Tarun ChelumallaGenAI against the slop: how to tell when output is yours
A short, mean checklist for spotting AI-generic before it ships under your brand.
Vaidehi KhuntOr skip the reading