CHAPTERS
- 0:02 – 0:32
Using AI to clear engineering backlogs without slowing growth
Siddhi frames the problem: Yhangry is at ~$15M GMV and targeting 10x growth, but engineering time gets pulled toward higher-ROI work. As a result, small but painful bugs linger and degrade the product experience.
- •High-growth context forces constant prioritization tradeoffs
- •Small “annoying” bugs are repeatedly deprioritized
- •Backlog accumulated during a busy period (mat leave timing)
- •Need for a way to fix low-priority issues fast without derailing roadmap
- 0:32 – 1:02
Building an autonomous bug-fixing agent in under four days
She describes shipping an autonomous bug fixer quickly and getting immediate impact: 25+ bugs fixed in the first week. She shares a rough benchmark for one-shot bug fixes and highlights the central challenge: giving the agent enough context to improve.
- •Autonomous bug fixer built in <4 days
- •25+ bugs fixed and shipped in week one
- •One-shot fix pass rate cited around 60–70%
- •Primary constraint is context: how to feed enough information for self-improvement
- 1:02 – 1:33
Turning “teach AI agents” into a growth and partnership wedge
Siddhi explains a non-obvious business use case: teaching AI agents in plain English as a differentiator. She connects this to Yhangry’s B2B vacation rental partnerships and uses education as a reason to get on conference stages.
- •Plain-English AI agent teaching is a competitive edge
- •Yhangry’s model includes both direct customers and vacation-rental partners
- •Pitch: teach ‘how to build AI agents in 30 minutes’ instead of product pitch
- •Education becomes a door-opener for partnerships and distribution
- 1:33 – 2:03
Free conference slots, then funnel attention into Yhangry’s product
She details the conference strategy: winning ~$50K worth of slots for free, then embedding Yhangry’s affiliate integration and AI product demo inside the “educational” deck. Attendees share slides and post on LinkedIn, compounding reach.
- •Secured ~$50K equivalent conference exposure at no cost
- •Deck includes affiliate link integration and Yhangry AI demo
- •Attendees screenshot slides and save them to use later (via Claude)
- •Post-event LinkedIn sharing creates strong organic amplification
- 2:03 – 2:33
Why the current chef-booking workflow feels outdated
Siddhi identifies friction in the existing product: too many steps and slow back-and-forth, causing drop-off. She notes Yhangry already has data about preferences and responsiveness, but historically lacked a way to match instantly and accurately.
- •Current customer flow has too many steps and delays
- •Back-and-forth communication causes conversion loss
- •Yhangry has rich data on chef/customer preferences and behavior
- •Core opportunity: instant, high-quality one-shot matching
- 2:33 – 3:03
Yhangry AI vision: ‘Claude for chefs’ + instant matching for customers
She lays out the product direction: an AI experience for customers to reduce friction and improve matching, and an AI assistant for chefs to eliminate repetitive admin. Early chef validation is strong, but it’s not yet good enough to ship.
- •Customer-side AI to generate fast, accurate chef matches
- •Chef-side AI to automate admin and repeated messaging
- •Positioning: ‘Claude for chefs’
- •Chef feedback/validation is positive, but quality bar not met for launch
- 3:03 – 3:34
Defining an MVP when chef workflows are highly divergent
Siddhi explains why the chef-side product is hard: chef needs and working styles vary widely. The team is trying to determine the minimal launchable version and how to handle open-ended requests.
- •Chef-side use cases vary dramatically between individuals
- •Hard to constrain requests into a predictable workflow
- •Need to choose an MVP that still delivers clear value
- •Balancing flexibility vs. reliability for first launch
- 3:34 – 4:04
Going company-wide AI-native: the personal and org reset
She shifts to how the organization is changing: AI agents and AI-native workflows across the company. She recounts an intense personal adoption curve, replacing leadership in engineering, and committing to an “all in” AI direction.
- •Company push for AI agents in every function
- •Rapid personal adoption of voice/Telegram AI workflows
- •Leadership change: fired tech lead, hired new head of engineering quickly
- •Explicit commitment to going “all in” on AI and rethinking org structure
- 4:04 – 4:34
Operationalizing agent adoption with weekly ‘agentic labs’
Siddhi discusses the training/enablement challenge: people claim they’re building agents, but proficiency varies widely. The solution is structured weekly sessions, hands-on diagramming, and using tools like Claude code to document workflows clearly.
- •Wide variance in employee AI/agent skill levels
- •Weekly agentic labs to standardize learning and output
- •Hands-on workflow mapping and shared understanding
- •Using transcripts + Claude code to create diagrams/documentation
- 4:34 – 4:55
Founder brand as a growth engine: teach what you know, funnel back
She closes with a growth insight: founder brand is about leaning into what you’re uniquely good at, gaining attention, and routing it to the business. Her advantage is translating technical AI concepts into plain English, informed by prior AI education.
- •Founder brand = do what you’re great at, earn attention, route to company
- •Her edge: simplifying AI concepts for broad audiences
- •Leverages prior AI academic background (Columbia, 2013)
- •Giving away valuable education for free drives demand back to Yhangry
