AI Is Unlocking Millions Of New Builders

Predictions and Expectations:

• Mukunda and Madhav expect that "more and more of these SaaS workflows are going to get consumed by an agent," predicting that SaaS companies must pivot to be "agent-first" to survive.
• The founders predict that "people are going to want to build more complex apps on the platform" as models become more capable.
• Mukunda expects the market to expand rather than shrink, believing "the rate of like software that you're expected to ship per week just keeps going up and up and up."
• The speakers believe that by the end of the year, there will be agents "running 24 hours and maybe hundreds of agents collaborating on just single tasks."
• Mukunda predicts that eventually "most of these models are going to get really commoditized," leading to similar behaviors and price competitiveness across providers.
• The founders expect that people who are "really close to problem domain expert" but blocked by technology will use Emergent to build things, creating a "camera explosion of ideas" that includes niche applications never built before.
• Mukunda believes that "the nature of software itself is changing," with "a lot more software will become agentic in nature."
• The speakers observe that "roughly 20%" of apps built on their platform are currently "agentic apps" where users embed the agent to power workflows.
• Mukunda expects that non-technical users will increasingly build "serious applications" rather than just prototypes, with 80% of current users being non-technical.
• The founders anticipate that "people who have had an idea for a long time... can now go and build things up," replacing the need for "dev shops" that charge "five hundred thousand dollars" with a self-service cost of "five thousand dollars."

Timelines and Milestones:

• Mukunda expects "by the end of the year" to see agents running for 24-hour horizons and collaborating in large swarms.
• The founders note that "every new model generation" presents a new opportunity, citing the progression from solving "JSON parsing" to enabling "extremely long horizon tasks."
• Madhav recalls applying to YC with the initial idea of automating software testing in "2023," noting it seemed "too crazy" to VCs at the time.
• The team pivoted to a general coding agent and began building it in "2024," after realizing the value of solving verification first.
• The company launched a small beta pro pilot in "June last year 2025" (contextually referring to the year prior to the 2025 launch timeline discussed, likely 2024 based on the "eight months since launch" metric relative to a 2025 interview, or potentially a transcript error for 2024; strictly quoting the text: "June uh last year 20 uh 25"). Correction based on context of "eight months since launch" and "2024" mentioned earlier for the pivot: The text says "June uh last year 20 uh 25" which implies the interview is late 2025, and the launch was June 2024.
• The founders observed that "Open-source is maybe three to six months behind" the leading proprietary models.
• The team notes that "Opus is a different class model right now," implying current capabilities for "extremely long horizon tasks."
• Mukunda mentions that "Model 4.6 is at 10 hours" for agent horizon, up from "4 hours" for Model 4.5.

Technology and Product Direction:

• The company is "experimenting with agent swarms" where "multiple agents can sort of coordinate on a single task" with an "overseeing agent" to monitor trajectories.
• Emergent is building "custom fine-tuned verification layers" to "augment" foundation models rather than competing directly with them.
• The platform utilizes a "multi-agent architecture" where a "driving agent" handles the main routine and "sub-agents" handle delegated tasks like testing or integration.
• The company plans to expand into features helping users with "distribution," "growth," and "how do you manage users," moving beyond just building the app.
• The founders are developing a "long-term memory" for agents that learns across sessions, allowing agents to auto-generate skills based on previous trajectories without re-prompting.
• The product strategy involves building "cloud sandboxes" on their own Kubernetes tech stack rather than outsourcing, to ensure the agent has the "same infra during the build time and the same infra during deploy time."
• The team is focusing on "reimagining the world" to allow agents to solve for "extremely long horizon tasks" enabled by models like Opus.
• The company is exploring "skills bench" and "continual learning" to improve agent performance without generating skills directly via the agent.

Market and Industry Outlook:

• The founders believe the market for "software engineering positions" is actually "going up" because "the more powerful the tools get, the more ideas you get."
• Mukunda expects the market to "expand" as "non-developers" are enabled to "not be developers," solving the problem of "agency" for individuals.
• The speakers predict a shift where "one person is doing, you know, like work of all three together" (PM, designer, engineer), combining roles that previously required a "five, six people team."
• The company sees a future where "every single country has an opportunity to build for a global audience," reducing the gap between developing nations and Silicon Valley.
• The founders expect that "70 80 percent" of their users will remain in "u.s europe" with "over 190 countries" represented globally.
• Mukunda believes the current "SaaS" model needs to change because "SaaS workflows are going to get consumed by an agent."
• The speakers anticipate a "hedonistic adaptation" where productivity increases as tools get more powerful, leading to higher expectations for shipped software.

Company Plans:

• The company is "hiring here in SF" for "research across the board, like backend engineers, frontend engineers," while maintaining a large team in Bangalore.
• The team plans to "reduce this gap between idea and reality" by enabling domain experts to build software without technical barriers.
• The company intends to build "production-ready software" from the ground up, including backend, frontend, security, and hosting, rather than just front-end prototyping.
• Emergent plans to support "multiple foundation models" to "extract more" capability than any single model can provide alone.
• The company is building "internal apps" using the platform itself, including a marketing CRM, customer support software, and a project management tool (Asana clone), to test complex scenarios.
• The founders plan to "keep delivering more and more things to our users" as the "spectrum keeps growing" on the side of post-building needs like growth and distribution.
• The team is focused on "building best verifiers" internally through "custom fine tuning" to ensure agent autonomy does not derail tasks.

Financial Guidance:

• The founders state that building a custom software solution previously would have cost "five hundred thousand dollars," whereas the new cost on Emergent is "five thousand dollars."
• The company mentions saving "three thousand four thousand dollars a month in subscription" by building their own internal tool instead of using Asana.
• A user recently "raised $4 million on an ad that was built on emergent," which the founders cite as an example of the platform's potential.

Risks and Caveats:

• Mukunda expresses concern about whether "entropic gonna eat everybody up" if models become too powerful.
• The founders acknowledge that "most of these apps they wanted to actually really build an app that works right" but notes that earlier competitors were "optimized for front-end prototyping" which may not be sufficient for production.
• The team notes that "non-technical users... even get panic as soon as they see a diff," requiring careful UI design to hide technical complexity.
• There is a risk that "if you start from a starting point, which sort of solves all of these problems... it's actually really hard to come from the other side and solve these problems," referring to the architectural difficulty of retrofitting full-stack capabilities.
• The founders mention the "Javan's paradox" where increased efficiency might lead to increased workloads rather than leisure, potentially creating burnout or "civil unrest" if not managed.
• There is a risk that "if you give your agents the same infra... you don't encounter those many problems," implying that not having control over infrastructure could be a significant hurdle for other platforms.
• The company notes that "skills cannot be generated by agent themselves" if done naively, requiring a CI/CD process to validate and add them to memory.

Confidence and Disagreement:

• The founders express "really confident about the product" being "head and shoulder above what what exists in the market today."
• Mukunda states "I definitely don't think" that software engineering jobs will be "gone," contrasting with the common fear.
• The speakers are "very confident" in their ability to "extract 23 more" (likely a transcription error for "30% more" or "a lot more") on top of foundation models through their harness.
• Mukunda believes that "a lot of people are building on emerent today" indicates a clear shift in how software is consumed.
• The founders are "very conscious" that "non-technical users probably don't know the concept of API key," so they built the experience to abstract this away completely.
• The team feels "really confident" in their "multi-agent system" and "long-term memory" architecture, having discovered these truths before they appeared in academic papers.