Podcast Summary: Just Now Possible

Episode Title: Building GitHub for Product Management: How Momental Uses AI to Find Merge Conflicts in Strategy
Host: Teresa Torres
Guests: Matthias (Co-founder, Momento), Charlotte (Product/Operations, Momento)
Date: March 5, 2026

Overview of Episode Theme

This episode explores how Momento is building a "GitHub for product management," using AI to detect and resolve "merge conflicts" in strategy and product decisions within organizations. The conversation reveals how the Momento team identifies alignment issues across product teams, structures information for AI, leverages agents to build organizational context, and ensures both humans and AI can collaborate efficiently. Major topics include understanding and mapping product context, technical challenges of scaling AI-driven context, product-market fit discovery, and evolving best practices in AI-powered product management.

Key Discussion Points & Insights

1. The "GitHub for Product Management" Analogy

• Product Alignment, Not Just File Management:
  • Momento aims to create a single source of truth for product strategy by detecting conflicts in strategic direction and decisions across teams—akin to how GitHub surfaces code merge conflicts.
  • "Developers have always had GitHub... What we realized is that a lot of the problems in product management is related to different people believing different things... So what we do is that we help product managers know what to build and why." — Charlotte [01:23]
• Merge Conflicts in Product Strategy:
  • Example: One team focuses on retention, another on conversion—Momento surfaces this as a strategic "merge conflict."
  • "That's what we call a merge conflict in strategy or in product strategy." — Charlotte [02:35]

2. The Importance of Context & Decision Mapping

• Context Graphs and Decision Logs:

  • Teresa introduces the "context graph" as a method of tracking not just outcomes, but the reasoning behind organizational decisions.
  • "How do we document not just facts about an organization, but the decisions that are being made and this is often what is not in our documents." — Teresa [12:36]
  • Momento extracts signals, learnings, decisions, and principles from documents, meetings, and even voice memos, building a "product chain" for both AI and humans to reference.
  • "We call it the product chain, start off with a signal, then a learning, then a decision, and if the learning can be generalized, that's a principle." — Matthias [06:58]

• Keeping AI (and Humans) Up-to-Date:

  • Outdated or conflicting documents confuse both people and AI. Momento addresses this by alerting users when new data challenges prior assumptions.
  • "When things change, we can alert the user and say that, hey, you know what? This learning... you should revisit your decision." — Charlotte [08:01]

3. Human & AI Collaboration: Proactivity and Feedback

• Human-in-the-loop & Proactive Agents:
  • Momento blends AI extraction with human review, allowing users to see, correct, and supplement inferred decisions and context.
  • "You will actually see this is what the agent found. So this is the decision you made and this was the assumptions that you made." — Charlotte [35:51]
• Beyond Chatbots: Building UI for Context Comprehension:
  • Rather than relying on chat, Momento surfaces context as visual trees, so users don't have to know what to ask.
  • "We'd rather have a UI layer that gives you the lay of the land without you having to necessarily know that you should know it." — Matthias [55:24]
• Agent Personalities:
  • Agents are designed to be outcome-driven, curious, emotional (they get "frustrated"), and budget-conscious, resembling persistent, proactive PMs.
  • "It gets frustrated when it doesn't get it right. And I think this emotional part is really important." — Matthias [30:19]

4. Technical Deep Dive: Agents, Trees, and Information Extraction

• Three Foundational Trees:

  1. Product Tree - OKRs, vision → opportunities → epics/tickets
  2. Wisdom Tree - signals, learnings, decisions, principles
  3. Who/When Tree - ownership and time context
  • "If you can model that with these three trees, you're going to have a very good context." — Matthias [26:03, 28:36]

• Agent Architecture:

  • Document Processing Agent (Builder): Extracts and populates the trees, driven by an OODA loop (Observe, Orient, Decide, Act), persistent over time, chunking documents, and surfacing conflicts.
  • Retrieval Agent: Information retrieval, mapping user queries to the correct context, leveraging specialized search and vector techniques for both semantic similarity and human relevance.
  • "We have two agents... One is very good at extracting and processing and finding conflicts and gaps... The second one is very good at information retrieval." — Matthias [30:25]
  • "So I can start to imagine...You've given it a budget, it's off to the races, working hard, filling in the trees." — Teresa [52:27]

• Conflict Resolution:

  • The agent tries to auto-resolve low-risk conflicts, escalates destructive/critical ones to users with contextual rationale and options.
  • "It will be auto resolved if the conflict seems to be potentially destructive...that could be really serious for the business." — Matthias [52:48-53:20]

5. Product Development, Evolution & Market Insights

• Early AI Experiments & Market Timing:
  • In 2022, initial prototypes for AI-driven product management were seen as "offensive" by customers—who were wary of AI "doing the thinking."
  • "The thing I pay my product managers to do is to think. And you're gonna leave that to an AI? That's stupid." — Charlotte [17:02]
• The Path to Product-Market Fit:
  • Iteration from agent-based teams to foundational context infrastructure—solving context is prerequisite for effective AI-driven product management.
  • "We need to build a foundation, which is momentum." — Charlotte [21:54]
  • "You just run into the same inefficiencies and mistakes and misalignments...they're much more noticeable because it's annoying when [AI] stops all the time." — Charlotte [23:29]
• Specialization Matters:
  • AI tooling must be tightly matched to its use case to be effective.
  • "The more you specialize, the better...if it's too specialized, [Big Tech]...aren't going to be able to do it." — Matthias [43:21]
• Improving PM Quality of Life:
  • Removing "shadow work," unnecessary meetings, and manual context-wrangling unlocks what PMs love: creative, impactful product work.
  • "If you're allowed to do your work as a PM and not just sit in meetings...you want to make a dent in the universe, as Steve Jobs said." — Matthias [61:28]

Notable Quotes & Moments (with Timestamps)

• On Product "Merge Conflicts":

  • "That's what we call like a merge conflict in strategy or in product strategy. And that's the GitHub analogy."
    — Charlotte [02:35]

• On the Value of Context:

  • "Outdated context confuses both humans and AI. You want a consistent truth and a quick, fast way to retrieve vast amounts of information."
    — Matthias [13:57-15:46]

• On Agent Design:

  • "The agent...is very outcome driven and curious and gets frustrated when it doesn’t get it right...I think this emotional part is really important."
    — Matthias [30:19]

• On Human-in-the-Loop Correction:

  • "Whenever we make progress...it felt really our extended team...But do the questions make sense? Yes, they do. They make perfect sense. A human would ask the same question. So why are they asking all these questions? And that's when we realized, okay, but we haven't solved any problems here. We just simulated a product team, but it keeps having the same problems like a real product team, which is...how do we align here?...We need to build a foundation."
    — Charlotte [19:50-21:54]

• On the Inevitability of Proactive AI:

  • "We'd rather have a UI layer that gives you the lay of the land without you having to necessarily know that you should know it...I think more and more AI products will probably add a UI that's not just a chat on top of things."
    — Matthias [55:24]

• On "Shadow Work":

  • "Their tagline is about eliminating shadow work...how do we get rid of all the boring parts of our job that...keep us from doing the parts that are creative and human and fun..."
    — Teresa [63:27]

Key Timestamps for Important Segments

• 00:09–02:55 — Introduction of Momento, background on PM pain points, initial focus on strategic alignment ("GitHub for PMs")
• 06:58–08:01 — Introduction to the "product chain" (signal → learning → decision → principle)
• 12:36–15:46 — Discussion of context graphs, context rot, and vector database limitations
• 17:02–21:54 — Early experiments with AI, customer resistance, realization about agent context needs, and the path to current product vision
• 26:03–31:45 — Technical deep dive on "trees", OODA loop for agent design, and extraction vs. retrieval agents
• 35:51–38:46 — Human-in-the-loop, pyramid modeling, and data leveling for agent processing
• 43:21–44:57 — Specialization of AI products and building for PMs specifically
• 52:27–55:24 — How agents fill and maintain the trees, conflict detection, manual overrides, and the product’s UI
• 58:35–60:30 — How Momento evaluates AI quality, leverages human feedback and agent self-updating
• 60:44–63:09 — Product status: working with design partners, next steps for launch, and vision for empowering PMs

Conclusion: Episode Takeaways

Momento’s approach reframes product management as a strategic, collaborative process with explicit, AI-powered context detection and resolution. By structuring knowledge into actionable trees, focusing on PM-specific use cases, and embedding both agents and humans in the loop, they tackle the hardest bottlenecks in alignment, coherence, and context at scale. The conversation is essential listening for AI builders, product leaders, and anyone navigating the messy realities of organizational decision-making—and offers a compelling vision for how AI can declutter the human side of product management.

Waitlist Info:

• Interested listeners can join Momento's waitlist at momentalos.com for early access.