How These 3 Founders are building on Open Claw | E2248

Episode Summary

Executive Summary: The episode explores how OpenClaw-style agentic AI is transforming startup operations by automating multi-step work across email, analytics, support, content, and code fixes. Founders argue these tools let small teams run multiple experiments, ship faster, and potentially lower startup budgets and staffing needs. Two builders demo practical use cases: a wellness founder uses agents as a chief-of-staff, and two developers show glasses and hosted-agent prototypes that bring AI into real-world workflows.

Main Topics: Agentic AI as a startup force multiplier (Priority: 5/5): Presh argues that AI agents can automate tasks, run parallel experiments, and let a small team behave like a much larger one. The core thesis is that startups are sets of hypotheses, and agentic tools let founders test several at once. OpenClaw as a digital chief of staff (Priority: 5/5): Presh demonstrates an assistant persona (AWA) connected to email, Telegram, PostHog, GitHub, and Slack that can gather user lists, draft docs, surface bugs, and prepare customer outreach with human approval. Real-world hardware integration via smart glasses (Priority: 4/5): Sean shows a prototype that pairs Meta Ray-Ban glasses with Gemini Live and OpenClaw to perceive the environment and execute tasks like adding items to an Amazon cart through voice and vision. Customer support and bug triage automation (Priority: 5/5): An inbox-monitoring agent identifies a support issue, correlates it with the codebase, suggests likely causes, and then Claude in Slack fixes the bug and creates a branch for testing. Startup economics and capital efficiency (Priority: 5/5): The conversation argues that AI reduces the cost and headcount required to build apps, which may shrink seed rounds, reduce dilution, and make bootstrapping or smaller raises more attractive. Hosting and productizing agent infrastructure (Priority: 3/5): Vishnu describes hosting OpenClaw in a VM, then turning the setup into a lightweight paid service because many users wanted an easier, safer way to run the agent.

Key Arguments: Agentic tools do not just make existing work faster; they enable entirely new workflows by giving models access to email, analytics, code, and other tools. A startup is fundamentally a bundle of experiments, and AI allows founders to run multiple theses simultaneously rather than sequentially. Many knowledge-work roles can be partially or fully automated into discrete tasks, reducing the need for large early-stage teams. Startup app companies may become cheaper to build and operate, but the number of apps may increase because experimentation is now affordable. Safety matters: once agents have real permissions, confirmation flows and sandboxing are necessary to prevent prompt injection or accidental misuse. Real-world interfaces like smart glasses could become a major entry point for agentic systems by turning vision and voice into actionable tasks. The same agentic stack can be applied across consumer products, support, content, inventory, and field operations, not just software demos.

Data Points: Consumer iOS apps shipped with AI tools: 3 - Presh says the team built and shipped three consumer iOS apps in the past year using AI tools. Time to complete a multi-step task: ~5 minutes - Examples included generating a user-reference document and pulling users from PostHog into a spreadsheet, then emailing them. OpenClaw adoption timeline: ~2.5 weeks - Presh first saw the tool about two and a half weeks before the episode. Task automation staffing comparison: 3–4 people instead of 10 - The hosts discuss how agentic tools could replace some roles and reduce team size for a startup. Budget and efficiency claim: 10x faster - Presh says a million-dollar round can now translate into roughly 10x faster product development. Cron schedule: 8 a.m. and 6 p.m. - Presh’s agent runs twice daily to search for relevant health and wellness news. Estimated VM hosting cost: $25/month - Vishnu says GCP medium instances were needed to run the agent reliably, costing about this much. Cheapest hosting option found: $7–$8 - Vishnu compared cloud providers and found a lower-cost host before productizing the setup. Current hosted service price: $3.99/month - Vishnu says agent37.com charges a few dollars a month to run a hosted OpenClaw instance. Daily operating cost target: $1.25/day - Vishnu estimates the hosting can be run for roughly this much if optimized properly. Active users on hosted service: 71 - Vishnu says the hosted version had around 71 active users and was still growing. Typical dilution per funding round: 10%–15% - Jason says founders often give up this amount each round in the modern venture market. Older dilution per round: 20%–30% - Jason contrasts current dilution with earlier venture norms.

Pivotal Quotes: "You can prove out multiple theses at the same time because of these Agentic tools." — Presh Dinesh Kumar: On how agentic AI changes startup experimentation and product strategy. "We basically took AI and we put it in a little closet. Now, you let the AI out of the dungeon and you say, Here's the keys to the kingdom." — Jason Calacanis: Explaining the shift from sandboxed LLMs to tool-using agents with real permissions. "I think we'll see more apps because apps are going to get so affordable to build and maintain and to improve." — Jason Calacanis: On the future of app counts, startup formation, and market fragmentation.

Implications: Agentic AI is pushing startups toward smaller teams, faster shipping, lower capital needs, and more automation across operations. It may both compress some roles and expand the total number of niche products and services built.