Mucker Capital went to New York Tech Week to host sessions for founders, operators, and builders thinking through how AI is changing product, growth, and search.
Among the sessions were:
NYTW Live Vibe-Coding Iron (AI) Chef Style Presentation
For this session, we gave three builders one real Mucker product challenge and asked them to prototype a solution live.
The builders were:
- Rock Vitale, Founder and CEO of Easie, an AI and automation consultancy for small and mid-sized businesses
- Elizabeth Lin, Design Program Manager at Ramp and Founder of Design’s a Party
- Zakir Jiwani, third-year undergrad at Drexel and campus leader for Livable, where he runs vibe-coding workshops
- Tony Yang, Head of Growth & Platform at Mucker Capital, hosted and ran commentary throughout the session
The prompt
The challenge: build an app or agent that could match Mucker founders with Mucker Operating Experts.
This was the prompt shown to the builders:
The current workflow:
If founders have a specific need or question to be answered that requires expertise, they can actively search the Mucker Operating Expert database within our private Mucker Founder Portal section of the website.
These queries can be specific to the industry or market they operate in, or related to a question that is not specific to their industry, such as advice on expanding to new regional markets like EMEA.
Once they identify experts through keyword search or filtering by industry or domain expertise, they can book a meeting with the identified expert.
The challenge is that founders often do not know what they do not know, and therefore do not actively take advantage of this operating expert resource.
In addition, the manual search and filtering capabilities are purely keyword-based, with no current ability to analyze context. As a result, the search does not always return the best-fit advisor, or the results page contains too many options to sort through.
The goal is to create a much more intelligent recommendation engine to match the right operating experts to founders.
The builders were given read access to two Airtable bases and asked to prototype a better way to make those matches.
Key Takeaways
Start with the agent, but do not skip planning.
The builders did not spend the whole session writing specs before touching a tool. They put the prompt into an agent, looked at what came back, and used that output to shape the next step.
Rock used Claude Code’s plan mode to work backwards from the desired result, map the user journey, and create an action plan before building.
Break the work into smaller pieces.
One practical takeaway was to avoid giving an agent one giant task.
Instead:
- Break the project into smaller chunks
- Treat those chunks almost like a Kanban board
- Keep the work in a GitHub repo from the start
That makes it easier to review changes, avoid bloated code, and roll back if the agent goes in the wrong direction.
Not every internal tool needs to become an app.
Elizabeth showed that the matchmaking problem might not need a full application at all.
By connecting Claude to Airtable and turning the instructions into a reusable skill, she created a version of the matching workflow that lived inside a chat window.
A founder could describe what they needed, and Claude could return relevant experts from the database.
Pretty and functional are still two different demos.
One prototype looked polished but was not fully wired up.
Another prototype worked more like a real matching engine, using embeddings, keyword ranking, and hard filters, but was not as visually polished.
Zakir summed up the tradeoff clearly: the version that works often does not look good, and the version that looks good often does not work yet.
The lesson is to use the visual prototype as a spec, then bring that design into the working version.
A demo is not the same as a product.
Rock’s prototype also showed what gets skipped when teams only optimize for speed. Semantic search can break in simple ways, like misunderstanding negation.
The last part of building still matters:
- Evaluation
- Security
- Authentication
- Database choice
- Hosting
- Real user testing
A working prototype is useful, but it is not production-ready just because it looks impressive in a room.
Future of Search: AI, SEO & AEO Panel
For our Future of Search: AI, SEO & AEO Panel, we brought together experts working directly on how AI is changing search, content strategy, and discoverability.
The Panelists
- Mike King, Founder and CEO of iPullRank
- Emily Richardson, Cofounder and COO of KNWN
- Alp Aysan, Cofounder of Cognizo and former search engineer on Microsoft Copilot
- Jason Patel, Cofounder and CEO of Open Forge AI
- Tony Yang, Head of Growth & Platform at Mucker Capital, moderated the panel
The conversation started with terminology: SEO, AEO, GEO, and the other labels emerging around AI search.
Mike pushed back on the naming debate and argued that the work is really about relevance engineering: understanding what signals help a system decide which sources to use in an answer.
Key Takeaways
AI search is not just Google search with a new interface.
Mike explained that AI search relies on retrieval-augmented generation.
A user asks a question. The system expands it into multiple synthetic queries, retrieves relevant passages, and then uses a language model to synthesize an answer.
That means the old playbook of ranking one page for one keyword is no longer enough.
Content needs to be structured for retrieval.
Emily explained that AI search does not read pages the same way Google does. Long “skyscraper” pages are less useful if the answer is buried.
The practical shift:
- Put important information near the top of the page
- Use question-based headings when appropriate
- Answer directly
- Use FAQs, short summaries, and clear sections
- Keep pages fast
The goal is to make it easier for AI systems to find and use the right passage.
Customer language should drive the content strategy.
Jason’s advice was simple: record customer calls, with permission, and look for patterns in how customers describe their problems.
One call a day gives you 30 calls a month. Those repeated questions and phrases can become the foundation for content that matches how real buyers ask for help.
He also noted that commercial content may be more valuable than broad informational content. That includes:
- Comparisons
- Buying guides
- How-to resources tied to a decision
The reason is simple: the buyer is closer to action.
Citations, mentions, and category relevance matter more than generic backlinks.
Mike was blunt that he does not care about backlinks in this environment the way traditional SEO has talked about them.
The panel focused instead on:
- Relevance
- Citations
- Mentions
A citation is when an AI answer links to a source. A mention is when the brand appears inside the answer itself.
Emily noted that mentions can be even more valuable because they show how the brand is understood across the web.
Jason added that the right source depends on the category. Some companies may need G2. Others may need Gartner, Reddit, YouTube, niche forums, or other third-party sources.
Busy founders should focus on practical, high-leverage actions.
When Tony asked each panelist what a busy founder should do, the answers were direct:
- Mike: create the definitive guide on the topic your customers care about
- Jason: fix structured data and schema markup
- Alp: build a realistic content calendar across blog, social, and third-party channels
- Emily: create content based on real experience, since AI cannot produce first-person knowledge on its own
Measurement needs to include brand visibility, not just referral traffic.
Mike suggested tracking AI search across three buckets:
- Performance
- Brand
- Input metrics
Referral traffic still matters, but it is only one piece.
Teams also need to understand share of voice, whether AI systems describe the brand accurately, crawl activity, and whether their content is relevant to the synthetic queries AI systems generate.
The panel also left founders with a bigger question: as AI does more of the synthesis, channels built on first-person knowledge may become even more important for discovery, including video, communities, forums, Substacks, and niche spaces.
