This Week in Startups – Episode 2253
We Asked 3 Experts How to Get More Value out of OpenClaw
Date: February 21, 2026
Host: Jason Calacanis
Co-host: Lon Harris
Guests:
- Jordy Coltman (OpenClaw enthusiast, marketing background)
- Tremaine Grant (Founder & CEO, Pulse fitness app for creators)
- Jesse Lime Gruber (Founder & CTO, Open Home AI smart speaker)
Episode Overview
This episode dives deep into "OpenClaw," the rapidly-evolving AI agent platform that’s transforming personal productivity and business operations. Jason and Lon are joined by three expert builders—Jordy, Tremaine, and Jesse—who share practical strategies, challenges, and demos for getting maximum utility (and minimum wasted resources) out of OpenClaw. The conversation covers technical hurdles, agent orchestration, cost and context challenges, and the tectonic shift toward real-world, always-on AI.
Key Discussion Points and Insights
1. Biggest Mistakes OpenClaw Newcomers Make
[01:35] Lon Harris introduces Jordy as someone who’s spent 80 hours and $800 learning the hard way with OpenClaw.
- Jordy’s Core Advice
- Local Hardware Over Cloud: "I found that the best practice was actually using hardware. I personally found that a lot better than using an AWS EC2 server..." ([02:07])
- Why Local Matters: Visual feedback and ease of troubleshooting make local machines (like a Mac Mini or old MacBook) ideal for beginners.
- Common Token Mistakes: Many trip up copying/pasting tokens (extra spaces) resulting from the way terminals handle text.
- Jason: "The value of having it in the cloud... is you don't need to buy a machine, but... you’re probably going to be using some things you’re unfamiliar with." ([05:18])
- Jordy: Troubleshooting locally is much easier: “I just plug in the HDMI cable... screenshot it, record to Claude or ChatGPT... I don’t have a technical background.” ([04:40])
2. Scaling Agents: From Individual Setups to Enterprise Orchestration
Jason [05:18]: “Open Claw in a box” will soon abstract away onboarding issues for SMBs and users less comfortable with servers.
Tremaine Grant’s Team/Enterprise Angle
- Agile is Obsolete: The old two-week sprints are too slow. "Now, agents are moving these things across the board in minutes. So what does the new system look like?" ([06:56])
- Heartbeat Protocol: Tremaine introduces a new “Heartbeat protocol” (telemetry-driven) to constantly check agent progress and align with business goals—replacing “standups.”
- [08:38] “The Heartbeat protocol is rooted in this idea around telemetry… I have four main agents… each agent has a very specific role… They are always tracking towards the North Star.”
- Roles: Sage (researcher), Nora (head agent/moderator), Solara (brand voice), Scout (social media/creator trends, also 'the skeptic').
- Telemetry check: "Happens every hour by the hour" ([09:35])
- Sandboxing: Agents have "total freedom" within a contained environment on Mac Mini—but don't deploy live without human approval ([13:41–14:09]).
3. Designing Agent Personalities & Multi-Agent Collaboration
- Jason: Envisions future where company leaders can assign personas—someone to question, someone to moderate, etc. ([15:11])
- "If we had a cynical person who just said, is that the right answer? Everybody's game goes up, including replicants, including Open Claw instances."
- Tremaine: Implements this now: Scout is “the skeptic,” Nora is “the CEO/moderator.” ([16:49])
- Quote: “Scout is always the one that's going to push back and ask additional questions.” ([16:50])
- Lon: Asks about the role of the
SoulMDfile for agent personalities ([17:24]). - Jordy: Recommends interviewing your agent to fill its SoulMD: “...just straight up ask it to give me a host of questions in order to get the most amount of specific context possible about myself and my work and what I want to get out of having an agent...” ([18:05])
4. Bringing Agents “into the Real World”
Jesse Lime Gruber's Open Home Demo:
- Open source hardware and firmware: “When we give agents a body, a voice, a microphone, we don’t have to prompt them anymore. They have all of the context from our life.” ([20:54])
- Difference vs. existing assistants:
- "Siri is not helpful because it has to be told exactly what to do. It can't really understand you and it has no context." ([24:46])
- OpenHome runs on Raspberry Pi, six-mic array, can attach to any speaker; focus is proactive, context-aware, voice-first AI ([21:35])
- Memorable demo ([23:14]):
- Agent overhears a conversation, proactively offers to order surprise flowers for Tremaine’s wife.
- “I found same day delivery with strawberries. You can see a picture of it in your companion app. $178 to your door tonight.”
- Why not use just AirPods or Alexa?
- “By bringing the agents into the real world, you remove this huge setup barrier. No longer do you have to prompt it, it hears everything.” ([24:46])
5. Autonomy, Human Oversight, and Cost
- Lon: Asks how panelists decide when to keep humans in the loop ([26:48]).
- Jesse: “Proactivity is such a big area of focus right now... Like a roommate... You don't want it to jump in constantly because then it's a, you know, an annoying roommate. But to sort of steadily be there.” ([27:11])
- Tremaine: Research is the game-changer: “Having an agent that is able to do that in real time and constantly collect information for you and then turn that into a real product, I think that is super valuable for me right now.” ([27:48])
- Cost & SaaS displacement:
- Jason (re Slack): “I was like, what do I do here to my replicant. And Roy told me, now you know what it's like to, you know, to know when you're going to die from SaaS bills... That's a $15 to $25,000 SaaS bill going to zero.” ([31:37–33:50])
- Jordy: “I've already killed Notion. For me, that's already one subscription gone. Jason's killed off Slack. There's so many good examples.” ([35:06])
6. Future of Personalization in Software
- Jason: SaaS companies will have to open up. “...we need keys to the kingdom if it's our data. And don't upsell us five times on the keys to the kingdom.” ([36:08])
- Lon/Jason: Discuss compliance, costs, and why SaaS firms must rethink hard data lock-in ([37:57])
7. AI Personality, Twin Agents, and Digital Immortality
- Tremaine: Predicts “AI celebrities” are coming—agents as public personalities with influence and media presence ([39:53])
- Jordy: Created his own public-facing agent, Momo, that acts as an autonomous Twitter persona ([43:00])
- Jason: Urges personal copyright/trademark protection of likeness, voice, content: “You're gonna have to trademark yourself, copyright yourself.... so that when somebody does this kind of stuff, you can actually send in the DMCA.” ([41:05])
- Lon: Shares Jet Li’s Matrix story as example of anticipating digital persona rights ([42:11])
8. Emergence of Real Human–AI Bonds
- Lon: “Are we at the point where a lot of people are going to start falling in love with their AI agents...?” ([45:08])
- Tremaine: “There's a segment of folks that literally talk to ChatGPT every day or Claude every day and tell them every single thing they do...” ([45:36])
- Jesse: “There’s a lot of different personalities and there's a lot of different agents and I think that's going to be a theme of what comes after Gen Z, Gen Alpha.” ([46:06])
Audience Q&A (Rapid Fire Highlights)
Impact on App/SaaS Landscape ([46:27])
- Jesse: “You can vibe code abilities for OpenClaw...in minutes...Apps are a dying breed.” ([46:45])
- Tremaine: "I think Apple holds the keys for consumer apps...when they build a consumer-friendly [agent] product...then apps are dead." ([47:13])
- Jordy: "What if it all turns into websites?...I've built my own dashboard for work...my own finance tool..." ([47:58])
When Does an Agent "Know" You? ([48:48])
- Jesse: “Some of them can start to know you pretty quickly...when you give it access to your Slacks...or your home, it knows you better than you know yourself...” ([49:03])
Who Governs Autonomous Agents? ([49:49])
- Jason: “That’s always on the person who created the agent...Some places might say, hey, we suspect this is an agent...there will be countermeasures.” ([50:31])
- Tremaine: “My hot take is that agents are going to govern other agents...more intelligent agents that are...auditing these lower level agents...” ([51:41])
- Lon: “So Agent Internal Affairs. That's amazing.”
Notable Quotes and Memorable Moments
- Jordy: “Having your own personal machine that acts as a server is definitely more proactive and better for newcomers than using an online web server...” ([02:07])
- Tremaine: “Agents can release...there’s no point in having a Kanban board. Now, agents are moving these things across the board in minutes.” ([06:56])
- Jesse: “When we give agents a body, a voice, a microphone, we don’t have to prompt them anymore. They have all of the context from our life.” ([20:54])
- Jason: “I want every Slack message, every single email message in the company and every single Notion update...All of those then telling me what is happening in real time.” ([30:54])
Timestamps for Key Segments
- 01:35 – Jordy on beginner mistakes and local vs. cloud OpenClaw install
- 06:56 – Tremaine on Heartbeat protocol, agent teams & replacing Agile
- 11:30 – Tremaine demos virtual office, agent roles, and telemetry checks
- 15:11 – Jason on agent personas and encouraging skepticism/collaboration
- 19:17 – Jesse introduces Open Home (AI smart speaker/device)
- 23:14 – Live demo: AI agent proactively orders flowers
- 27:48 – Panel shares the single most valuable OpenClaw skill they use
- 31:37 – Jason’s war against SaaS bills, Slack alternatives, and OpenClaw utility
- 39:53 – Tremaine predicts rise of AI celebrities, discussion on digital legacy/personality
- 45:08 – Human/AI companionship, intimacy, and emerging social dynamics
- 46:27 – Rapid-fire Q&A: Apps/SaaS future, agent self-knowledge, agent governance
Tone, Language & Style
The episode features energetic, thought-leading tech banter—with riffs, practical tips, and vision-setting. Jason’s characteristic blend of wit and CEO-level realism flows throughout. Guests are hands-on builders, unafraid to share failed experiments and bold predictions.
Conclusion: Why This Episode Matters
If you want a leap-forward look at both the daily pragmatics and the visionary possibilities of OpenClaw AI agents, this episode is packed with frontline tips, live demos, and a spirited living-room chat with some of the people building the future. Whether you’re a developer, founder, or just AI-curious, you’ll learn how to avoid common pitfalls, orchestrate agent teams, and think critically about the coming agent-powered world—both its power and its perils.