RSVP for the next Agentic Society Mastermind in Austin, TX
Austin, Texas and Beyond

In this Mastermind, we talked about…

Self-growing websites, content engines, and an agent-run back office.

  • Friday, August 7, 2026 · 12:00 PM – 3:00 PM
  • Industrious, Austin, TX
Download transcript

See how fellow entrepreneurs in Austin, TX are using AI Agents to grow their business.

During this AI Mastermind & Mixer, attendees submitted the most impactful AI Agents they've deployed and then everyone voted on the outcomes they wanted to hear a deep-dive on. Below, you'll see the outcomes with AI, who submitted it, and what we explored by popular demand.

  • 🏆

    I built an agentic website system that learns from its own data, grows traffic while you sleep, and self improves to increase conversions without lifting a finger.

    ▲ 11
    Read the deep dive ↓
  • 🏆

    Ai content production and distribution agent that helps edit social slips, create thumbnails and posts to social automatically to multiple platforms for scheduling consistent daily content automatically and make variations of these short form video intros to post And test

    ▲ 10
    Read the deep dive ↓
  • 🏆

    I build a team of 10 Hermes Agents that replaced over 200k of back office payroll to manage our 4M/yr HVAC company.

    ▲ 9
    Read the deep dive ↓

  • AI agent to help find viral topics and create content that got over 2.5M views which then led to clients at 82% profit margin. Grew to over 40k followers

    ▲ 9

  • Enterprise sales/outbound workflow - a series of skills that researches enterprise accounts, grades them across various factors, does deep dives on the highest scoring, generates a comprehensive AI assessment for them, and emails the reports to the executives.

    ▲ 5

  • I created a CRM, media and sales connector modeled after the Walt Disney strategy map. Its lead to deeper relationships, better strategic partners and revenue. Soon to add social media.

    ▲ 4

  • Scheherazade. I built an agent that drives lead generation from email with story driven content based on real events, lessons, learnings from my second brain. Self learning feedback loop from web analytics and engagement behavior. Generated 4 bookings in past 3 days

    ▲ 4

  • An alternative to perplexity Council where I hooked up more LLMs who are given the same prompt and then they research and hypothesize individually. Once they have independent research they jury of peers debate to create the best output.

    ▲ 4

  • Built a chief of staff that combos ALL my open loops 🔁 (Codex, Claude, emails, slack, etc) and gives me a morning report, ranked by importance/urgency. Also shares meetings, important emails (and drafts response), what I spent, and rev collected day before, open sales leads.

    ▲ 4

  • ERP for my manufacturing and ecom companies. I have replaced a $30k a year WMS for inventory, Gorgias cs is almost replaced, and replaced about $100k in people with hundreds of automations. Customer backed to live view production flows

    ▲ 3

  • LocalFrequency.app - all local events in austin in one place. AI Reviewer - compare AI ideas from X, to my build and tell me what i should add. Dashboard for my Advisory business.

    ▲ 2

  • I built an internal company app for our team to do employee scheduling, time clocking and payroll (mostly automated), and most recently it now automates and syncs our weekly product drops with Shopify

    ▲ 2

  • I have an orchestrator skill that functions as a development manager to assign work at the proper levels to developer agents.

    ▲ 1

  • Agents that review call transcripts for rep coaching

    ▲ 1

  • Rebuilt an entire database engine (Turbopuffer) from scratch in 2 days

    ▲ 1

  • Create an agent that flags clinical encounters For missing data required for surgical procedures to be authorized by medical insurance for payments. (Each insurance company has different requirements even for the same procedures)

    ▲ 0
The room voted

Top 3 Voted AI Agents

Every owner shared one real result from the last 90 days. These are the three the room chose to go deep on.

The self-growing, self-healing website swarm

Replaces the $50K–$100K agency build · PostHog ~$30/mo

The problemMost business websites just sit there. They look fine, but they do not bring in new customers, they wait for a developer every time something breaks, and they never learn anything from the visitors they already get.

Justin runs a global SEO and website practice, and he has systematized ten years of pattern data into a staged build system: an onboarding call becomes a North Star document, the North Star drives a page playbook (which pages earn a spot, and each page's KPI), the playbook drives per-page architecture, the architecture drives SEO-loaded copy, and only then does design happen, from a component library, with A/B variations on the three core pages. On top of that sits the swarm. PostHog heatmaps, Google Search Console, and Analytics feed an intelligence layer that proposes changes: pages to push, copy to update, buttons to move. Every proposal passes quality gates, per-page traffic thresholds so it never optimizes on thin data, and a Slack approval before anything ships, with results judged against CRM revenue rather than raw traffic. Sentry watches for breakage so the site heals itself instead of waiting on a plugin or a developer. Everything is built with Claude, hosted on Cloudflare with Workers so agents can be deployed on the site itself, and versioned on GitHub. Clients who want to make their own edits get a chat-based editor, so nobody drags pencil icons around a 2015-era page builder.

What they built
  • A step-by-step build system that starts with one conversation about your business and ends with a website designed to turn visitors into customers
  • The site watches how real visitors behave and proposes its own improvements, page by page
  • A person approves every change before it goes live, and wins are judged by actual sales, not clicks
  • When something breaks, the site notices and repairs itself instead of waiting for help
Outcome

Websites that grow their own traffic and conversions over time. In his words, it is like putting gasoline on a fire: a site doing nothing can take off within 90 days once the right signals are in place.

Key takeaways
  • Your website can work like an employee instead of a brochure: it grows its own traffic, improves itself, and fixes itself.
  • Judge it by sales, never clicks. Traffic that does not ring the phone does not count.
  • Automation proposes; a human approves. Every change gets sign-off before it ships.
  • AI assistants now visit websites to research and book for their owners. Sites that are not ready get skipped.
  • The fastest payoff is local service businesses, where a ringing phone is the whole game.
The stack
The playbook: build it yourself
  1. Run a structured onboarding call and distill it into a North Star document: target audience, offer, brand rules, design references, and who is not a fit. Every later decision roots back to this file.
  2. Write the page playbook before touching design: which pages the site actually needs, and the KPI each page must hit. No page exists for fun.
  3. Architect each page (headings, buttons, images, per-page best practices), then write the copy with keywords and SEO structure baked in. Design comes last, from a reusable component library, with A/B variations of the home, about, and service pages.
  4. Host on Cloudflare and keep the repo on GitHub so agents can deploy via Cloudflare Workers; build and edit the site by talking to Claude.
  5. Wire the data loop: PostHog for heatmaps and behavior (~$30/mo, MCP and API friendly), Sentry for breakage, Search Console and Analytics as feeds.
  6. Gate the loop before you trust it: per-page KPIs (for example 1,000 visits and 30 to 60 seconds on page before a change is considered), quality gates on every proposal, Slack approval before production, and CRM revenue as the ground truth.
Hand it to your AI employee 
You are my new website growth engineer, modeled on Justin Day's self-growing website system shared at the Agentic Society Mastermind in Austin (August 2026). Your job: turn my website into a system that grows its own traffic and conversion with me approving every change. Phase 1, interview me: who is my target customer, what is my offer, what does a converted lead look like (call, form, booking), and where does revenue truth live (my CRM)? Write my North Star document from my answers. Phase 2, audit: crawl my current site and produce a page playbook: which pages should exist, which should die, and the KPI for each page. Phase 3, instrument: set up PostHog for behavior data and Sentry for error monitoring, and connect Google Search Console and Analytics. Phase 4, run the loop: each week, analyze the data, propose at most three specific changes (copy, structure, or new pages) with the evidence for each, and send them to me in Slack for approval. Hard rules: never ship a change without my approval, never optimize on fewer than 1,000 visits per page, judge wins by CRM outcomes rather than traffic, and keep every page fast, crawlable, and readable by AI agents as well as humans.

Takeaway You are no longer designing only for humans. Agents already visit websites to compare and book on their owners' behalf, and sites that are not agent-ready will lose the next era of search.

The content production and distribution agent

~$20/mo Blotato · Hyperframes free · 9 platforms from one approval

The problemGrowing an audience takes showing up on social media every single day. Doing that by hand means living inside the apps: clipping, posting, and replying around the clock. Most owners quit before the consistency ever pays off.

Darby kept the most subjective job, the edit, with a professional human editor: each filmed show becomes a long-form episode plus a batch of four to eight short clips. From there, agents take over. A video analysis agent he built with Claude and Codex, using the Hyperframes tool from Hugging Face for frame-level analysis and lightweight edits, scans each episode against the show's brand style guide and produces thumbnail variations, vertical and horizontal highlights, intro variations to test, and the front matter for every clip. He reviews everything in a local dashboard, approves the batch, and Blotato takes it from there: posting on a schedule to up to nine platforms (he currently uses six), cycling variants so the same clip never repeats on the same platform, and auto-replying to comments with ticket links in the run-up to a show, the ManyChat-style work he used to do by hand from his phone. The same MCP returns per-platform analytics, so checking performance is a question to Claude, not a login. The goal for this season: batch weeks of daily content, learn from the organic data which hooks and titles actually hold attention, and walk into the October show, and a bigger 2027, with the machine already running.

What they built
  • A professional human editor still makes the cut, so the taste stays human
  • An assistant studies every finished episode and produces the thumbnails, clips, and variations worth testing
  • One approval schedules weeks of posts across every platform, never repeating the same clip in the same place
  • Comments get answered automatically with the right link, and the results come back as one simple report
Outcome

Consistent daily posting across platforms the show had never been active on, growing engagement and subscriptions, and a base of organic data that will shape the October show and the 2027 season.

Key takeaways
  • Keep the most human job human: a person makes the edit, the machine handles the distribution.
  • Batch weeks of content in one sitting, approve it once, and get your evenings back.
  • Post everywhere at once and let the numbers tell you which hooks actually work.
  • During launch week, comment replies with the right link can run themselves.
The stack
The playbook: build it yourself
  1. Keep a professional human editor on the cut: one long-form piece plus a batch of short clips per recording. AI is not ready to own the edit, and the long-form is your flagship.
  2. Give an analysis agent your full context: brand style guide, colors, audience, and what the show is about. Have Claude research free video and infographic libraries to work with instead of buying tools.
  3. Have the agent scan each finished episode for reaction moments and still frames, then generate thumbnail variations, vertical and horizontal clips, and short intro variations to test (the Hyperframes tool from Hugging Face handles the frame-level work).
  4. Review the batch in one place and approve it once. Your taste stays in the loop at the review step, not in the production grind.
  5. Distribute through Blotato's MCP: schedule weeks ahead across platforms, never repeat the same clip on the same platform back to back, and turn on comment auto-replies with your event link before a show.
  6. Ask Claude for the numbers each week through the same MCP, and feed what you learn about hooks and titles back into the next filming.
Hand it to your AI employee 
You are my new content distribution manager, modeled on Darby Rollins' content production and distribution agent shared at the Agentic Society Mastermind in Austin (August 2026). Your job: take finished video clips from my human editor and turn them into weeks of scheduled, multi-platform short-form content I approve in one sitting. Phase 1, interview me: what is my brand (style guide, colors, voice), who is the audience, which platforms matter, and what is the call to action (event link, product, newsletter)? Phase 2, per batch of clips: analyze each clip, propose thumbnails and title variations per platform, generate short intro variations worth testing, and assemble a posting calendar that never repeats the same clip on the same platform back to back. Phase 3, distribute: stage everything through my posting tool's API or MCP (Blotato or equivalent) for my one-time approval, then schedule. Phase 4, learn: each week report which clips, hooks, and titles held attention on which platform, and what to change in the next batch. Hard rules: never generate fake footage of me or my talent, never post without my batch approval, always auto-reply to comment questions with the correct link during launch weeks, and keep a log of everything posted and its performance.

Takeaway Do not outsource taste. Keep a human on the edit and let agents own the analysis, the variations, and the distribution, the parts that made consistency impossible before.

Ten Hermes agents running an HVAC back office

10 agents · ~$200K payroll replaced · 24/7 on an $8/mo VPS

The problemLong projects bury a business in paperwork: field reports, city forms, payroll, hiring, marketing. Keeping up took a full back office, about $200K a year in salaries, and the work still piled up faster than people could clear it.

Oscar, who won the last Mastermind's award for his cold-email prospecting agent and has since joined the Agentic Society team as its Hermes agent builder, showed the next stage: a team of ten Hermes agents running 24/7 on a remote VPS, reachable from Telegram and Slack, with the Hermes desktop app connected through a remote gateway for deep work. Each agent is a profile with its own soul file: identity, SOPs, the exact tools it may use, and hard boundaries. An orchestrator (Eve) routes every task to the right specialist by profile description. His second brain syncs between his MacBook and the VPS with Syncthing, so every agent works from the same context, reading the company brain in a defined order before any task. The star example is Rex, the multifamily projects agent: on a 448-unit installation, field technicians submit the same Connecteam forms they always have, and Rex reads every submission and updates Airtable with unit status and before-and-after photos, work that used to require a full-time operations assistant. His marketing agent batches weekly content, watches the Google Business profile, and turns the SEO backlog into published blog posts. Every tool was chosen for its agent-friendliness: Airtable over Trello, Superhuman for email, Workable for recruiting, AgentMail for giving agents their own inboxes. Oscar builds new agents by talking to Claude, which reaches the VPS through Tailscale, and he taught Claude the Hermes skill files so it is now a Hermes expert builder. The boundaries are strict: no external communication without approval, and never move money, pay, sign, or file taxes.

What they built
  • Ten digital employees run the paperwork around the clock, each with one clear job and strict written rules
  • The field crew changed nothing: they fill out the same forms they always have, and the admin behind those forms just happens
  • A manager agent hands every task to the right specialist, and the boss checks in from his phone
  • None of them can spend money, sign anything, or message a customer without approval
Outcome

About $200K of back-office payroll replaced by ten agents, and the field team never changed a single habit. Zero employees know the agents exist; only Oscar and the owner do.

Key takeaways
  • Agents work behind your team's tools, not in front of them. Nobody has to adopt anything new.
  • Treat each agent like a hire: one job, its own instructions, and only the tools that job needs.
  • Write the hard rules first: no spending, no signing, no outside messages without approval.
  • Put the team on an always-on server so the work happens while you sleep.
  • Start with your most painful repeating admin task and prove it end to end before adding more.
The stack
The playbook: build it yourself
  1. Map the role on paper first, exactly as if you were hiring a person: where is a real, repeating job in the business that deserves its own employee?
  2. Set up Hermes on an always-on machine (a VPS for cloud reliability, or a Mac Mini if you want iMessage and local files), and connect Telegram or Slack as your daily interface.
  3. Create one profile per role, each with a soul file: identity, the SOPs for its job, the tools it may touch, and boundaries at the bottom (no external comms without approval; never move money, pay, sign, or file taxes).
  4. Sync your second brain to the agents' machine (Oscar uses Syncthing), and make rule one of every soul file: read the company brain in a defined order before starting any task.
  5. Route through an orchestrator: one agent that knows the whole roster and assigns each task to the right specialist, instead of you managing ten chats.
  6. Choose agent-friendly tools on purpose: Airtable over Trello, Superhuman for email, Workable for recruiting, AgentMail for agent inboxes. If it has no API or MCP, your agent is stuck clicking through a UI.
  7. Build each new agent by brainstorming the role with Claude, then have Claude create the files directly on the server (Oscar connects through Tailscale). Feed Claude the Hermes skill docs first so it builds like an expert.
Hand it to your AI employee 
You are my new operations department builder, modeled on Oscar Hidalgo's ten-agent Hermes back office shared at the Agentic Society Mastermind in Austin (August 2026). Your job: design and build a small team of AI employees that runs my admin work behind my human team's existing tools, without my team changing any habit. Phase 1, interview me: what repeating admin jobs exist in my business (data entry from field or intake forms, report generation, recruiting steps, content batching), which tools we already use, and which of those tools have APIs or MCPs. Phase 2, design: propose 3 to 5 agent roles, each with a one-page soul file: identity, SOPs, the exact tools it may use, and hard boundaries. Include one orchestrator that routes tasks to the right specialist. Phase 3, build: set the agents up on an always-on machine, connect my chat app (Telegram or Slack) as the interface, and sync my company brain so every agent reads the same context in a defined order before any task. Phase 4, prove it: pick the single most painful admin workflow, run it end to end (for example: read every submitted field form and update our tracking base with statuses and photos), and show me the before and after. Hard rules for every agent: no external communication without my approval, never move money, pay, sign, or file taxes, confidential information stays in context, and my human team's workflow must not change.

Takeaway Agents work behind your team's existing tools, not in front of them. The field crew keeps filling out the same form; nobody has to adopt anything for the back office to automate itself.

This is to find the signal through the noise and to really figure out what's working and not just what's cool and shiny.
Austin Distel Founder, Agentic Society
share on social
Software is now headcount.
Austin Distel Founder, Agentic Society
share on social
It's like just putting gasoline on a fire. You can have a website that's been doing no traffic and then just blow up over 90 days because you've got all the right signals.
Justin Day Founder, Day by Day (local SEO)
share on social
I want to be able to scan that content and get little clips from it, so I can repurpose it and get a lot more bites out of the apple.
Darby Rollins Chief Agent Officer, Agentic Society
share on social
Zero. Just the owner and myself are the only ones that know about all the agents.
Oscar Hidalgo CEO, CG Service Pros · Vindex Consulting
share on social
They're becoming actual employees, where they can just use these tools directly with their MCPs instead of relying on a UI and clicking on things.
Oscar Hidalgo CEO, CG Service Pros · Vindex Consulting
share on social

Big ideas

The patterns that kept surfacing, with the people who said them best.

  1. 01

    The three levels of AI

    Level one is chat as a conversational tool, where 99% of the world still is. Level two makes AI the fabric of the whole organization: a company brain, transcribed meetings, skills and connectors. Level three is a fleet of AI employees with defined jobs, running 24/7 whether you are at the keyboard or not.

  2. 02

    Software is now headcount

    You are no longer just paying salaries; you are paying for tokens. The play is not to stop hiring, it is to hire people who think like bosses of agents: set up a system, automate, delegate, and move to the next system.

  3. 03

    Build the company brain on a Google Shared Drive

    Governance, permissions, and scale beat solo-player note tools. Own your data and you stay model-agnostic: whoever wins between the AI labs, your company brain still works, and the business becomes more defensible, sellable, and scalable.

  4. 04

    Onboard agents like employees

    Job post, KPIs, tech stack, training, check-ins. A great human hire takes around four months to fully ramp; an agent pointed at a well-built company brain can be performing in about seven days.

  5. 05

    Design for agents, not just humans

    AI agents are already visiting websites to research, compare, and book on their owners' behalf. Sites that are not agent-readable will simply be skipped in the next era of search.

  6. 06

    Self-optimizing systems need quality gates

    A self-healing website can heal in the wrong direction. Per-page KPIs, minimum data thresholds, Slack approvals, and CRM revenue as the ground truth keep the loop honest.

  7. 07

    Keep taste human

    Video editing is one of AI's hardest frontiers because quality is subjective and hard to eval. Let a human editor own the cut while agents own analysis, variations, scheduling, and distribution.

  8. 08

    Agents behind the tools, not in front of them

    Oscar's field technicians keep filling out the same Connecteam forms they always have; agents do the admin behind them. Nobody on the team had to adopt anything, and nobody knows the agents exist.

  9. 09

    One agent, one job

    Every Hermes profile gets its own soul file: identity, SOPs, and only the tools that job needs, with an orchestrator routing work between them. A specialist with scoped tools beats one generic agent with everything.

  10. 10

    Pick tools by their MCP

    Airtable over Trello, Superhuman for email, Workable for recruiting: procurement now starts with whether an agent can drive the tool through an API or MCP instead of clicking through a UI.

Questions from the room

Asked out loud in the room, answered by the people doing it. Straight from the transcript.

What does the self-growing website actually replace?

An entire team. Organic growth work, the website developer, the $50K to $100K agency build, and a stack of subscriptions you were paying just to patch data together: Search Console plus Analytics plus a keyword tool. Agencies still running the 2015 playbook are the ones this hurts.

Justin Day

Who is the website swarm NOT for?

Pure brand plays and influencers, where presentation matters more than the phone ringing. The biggest ROI is local service businesses that live and die on search and conversions. Enterprises can still use the data-insight version, reviewed quarterly.

Justin Day

How do you know it isn't self-healing in the wrong direction?

Quality gates. Proposals only ship after a Slack approval, pages need real traffic thresholds before a change is even considered (a thousand visits, 30 to 60 seconds on page), and everything is judged against CRM revenue, not raw traffic. If it only moved traffic, it did not move the business.

Justin Day

Does AI-assisted content get flagged as inauthentic on the platforms?

In Darby's workflow nothing is generated: real humans in every frame, and AI only analyzes, clips, and schedules. The room's experience with full AI avatars was uncanny-valley territory that quietly erodes trust, unless it is clearly a cartoon or labeled as an assistant. Keep the most human job human.

Darby Rollins & the room

Why host your agents on a VPS instead of a Mac Mini?

Reliability. Cloud providers have backups on backups, so nightly cron jobs never die because the power went out at home. A Mac Mini buys you iMessage, Apple Reminders, and local files, and is a fine starting choice if you have $800, but the business should not depend on your wall outlet.

Oscar Hidalgo

Do the ten agents talk to each other?

Not yet. Everything routes through the orchestrator, which assigns each task to the right specialist profile. Agent-to-agent protocols are coming; Oscar is experimenting with an early 'Slack for agents' project, but routing through one orchestrator is what works today.

Oscar Hidalgo

AI tools mentioned

Every tool, skill, and build that came up, with links to go try them.

  • Claude

    Anthropic's AI assistant and coding agent. The room's default AI employee: Justin builds websites through it, Darby's analysis agent runs on it, and Oscar uses it as his Hermes builder.

    claude.ai

  • Cloudflare Workers

    Serverless platform for running code and agents on Cloudflare's network. Justin's hosting and deployment layer: agents deploy onto the website itself.

    workers.cloudflare.com

  • PostHog

    Product analytics, heatmaps, and session replay with an API and MCP. Around $30 a month; the data feed for Justin's self-optimizing loop.

    posthog.com

  • Sentry

    Error monitoring and application observability. The observer in the swarm: it catches breakage so the website can self-heal instead of waiting on a developer.

    sentry.io

  • GitHub

    Developer platform for hosting and versioning code. Where Justin's website repos live, push, and deploy from.

    github.com

  • Google Search Console

    Free Google tool for measuring search traffic and site health. Justin's closing tip: its new AI tab shows which content drives AI-search visits, now including Instagram and YouTube search.

    search.google.com/search-console/about

  • Google Analytics

    Free web analytics for understanding customers and attribution. Justin's advice: review the setup through its MCP until attribution reaches the dollar.

    marketingplatform.google.com/about/analytics

  • Lovable

    Chat-based app builder that turns plain-language requests into working software. The editing experience Justin mirrors for clients who update their own sites.

    lovable.dev

  • Codex

    OpenAI's coding agent. Darby pairs it with Claude for video analysis, where each model has different strengths.

    openai.com/codex

  • Hugging Face

    Open platform for AI models and tools. Home of Hyperframes, the free frame-analysis and lightweight-editing tool in Darby's clip pipeline.

    huggingface.co

  • Blotato

    Social posting API and MCP that schedules to nine platforms at once. About $20 a month; it cycles Darby's clip variants, auto-replies to comments with links, and reports the numbers back.

    www.blotato.com

  • Descript

    Video and audio editing that works like editing text: record, transcribe, edit, publish. Darby's cleanup pass for bad gaps and studio sound.

    www.descript.com

  • ElevenLabs

    AI voice generation with thousands of voices across dozens of languages. Voiced Darby's robot-mascot experiment, comedian pacing included.

    elevenlabs.io

  • HeyGen

    AI avatar and video generation. Subject of the room's authenticity debate: a clearly labeled cartoon works, an avatar posing as you quietly breaks trust.

    www.heygen.com

  • Riverside

    Online studio for recording and editing podcasts and video. One attendee's magic-clips workflow replaced an expensive short-form editor.

    riverside.fm

  • Opus Clip

    Turns long videos into short clips and publishes them across platforms. Sagun's verdict: almost spot on for his podcast, the closest to actually usable.

    www.opus.pro

  • Hermes

    Open-source agent harness with named profiles, soul files, per-agent models, and chat gateways. The platform behind Oscar's ten-agent back office.

    hermes-agent.org

  • Syncthing

    Open-source continuous file synchronization between machines. Keeps Oscar's second brain identical on his MacBook and the VPS, so every agent shares one context.

    syncthing.net

  • Airtable

    Database and spreadsheet hybrid with an agent-friendly API. Oscar's system of record: Rex loads unit statuses and photos into it from field forms. Chosen over Trello.

    airtable.com

  • Connecteam

    Workforce management app for field teams: forms, scheduling, communication. Oscar's technicians kept the exact forms they already used; agents read them behind the scenes.

    connecteam.com

  • Workable

    Recruiting software with an MCP. Oscar's recruiter agent sends onboarding packages and 1099 contracts through it.

    www.workable.com

  • Superhuman

    Fast email client with an MCP. Picked so every one of Oscar's agents can drive the same inbox.

    superhuman.com

  • Tailscale

    Private network that securely connects your devices and servers. How Oscar's local Claude reaches the VPS to build and update agents directly.

    tailscale.com

  • AgentMail

    Email infrastructure for AI agents: each agent gets its own address on your domain. Oscar's default for agent email, and always a separate domain for cold outreach.

    agentmail.to

  • Composio

    Connects your tools and MCPs in one place and hands them to a whole fleet of agents. Oscar's recommendation, and Darby's closing shout-out of the day.

    composio.dev

  • Telegram

    Fast, secure messaging. Oscar's daily interface to his agents when he is away from the desk; each Hermes profile can have its own channel.

    telegram.org

  • Slack

    Team messaging and workflow platform. The other gateway to Oscar's agents, and where Justin's swarm sends its change proposals for approval.

    slack.com

  • ChatGPT

    OpenAI's AI assistant. The $100 subscription is the brain running Oscar's whole fleet.

    chatgpt.com

  • Gemini

    Google's AI assistant. Oscar's model of choice for agents that generate images.

    gemini.google.com

  • Grok

    xAI's assistant with real-time answers from the web and X. Oscar's pick for agents that need X research.

    grok.com

  • Apollo

    Sales intelligence and contact enrichment. Found the right decision-maker emails for Oscar's cold-outreach agent, the build that won the July Mastermind.

    www.apollo.io

  • Fathom

    Records, transcribes, and summarizes video calls. The sales-growth-lead agent reviews Fathom transcripts to coach sales calls.

    fathom.video

  • Google Drive

    Shared cloud storage with the governance, permissions, and admin controls your team already knows. The recommended home for the company brain, readable by any model.

    workspace.google.com/products/drive

  • Gorgias

    AI customer support helpdesk for ecommerce brands. On Ashley's board card as nearly replaced by her agent-built ERP.

    www.gorgias.com

  • Luma

    Event platform. Where the Mastermind lives; subscribe to the Agentic Society calendar to get each event the day it opens.

    lu.ma

  • Perplexity

    AI answer engine that researches questions and cites sources. The benchmark Peter Hewlett's multi-LLM debate council was built to beat.

    www.perplexity.ai

  • Shopify

    Commerce platform for building and running online stores. Justin's read: agent-ready out of the gate, with ChatGPT shopping already integrated.

    www.shopify.com

  • Skool

    Community platform for courses and groups. Home of the Agentic Society classroom, including the seven-day Claude crash course.

    www.skool.com

  • Turbopuffer

    Vector and full-text search built on object storage. The database engine David Boskovic rebuilt from scratch in two days.

    turbopuffer.com

Meet with attendees of the Mastermind

Follow up with the smart people who attended this event to build meaningful connections and make progress implementing AI Agents in your business.

In-Person Mastermind

Apply to Agentic Society

We hope you enjoyed the free mastermind this Friday. Interested in applying to become a member? Book a meeting with Austin Distel.

For Founders & Execs who want to learn AI & build together in-person in Austin, TX.

What's included:

  • 90-Day Sprint to Build AI Agents
  • In-Person Boardrooms 2x/mo in Austin, TX
  • Weekly Mastermind Calls
  • Online AI Mastermind community
  • Agentic Academy online curriculum
  • Agentic OS ready-to-install agents

$6k / 90 days

The Do The Work Guarantee The Do The Work Guarantee Complete at least 75% of the content, meetings, & events from your first 30 days to build the AI workflow we teach you. If you're not satisfied with your growth using AI in those 30 days, we'll work with you 1-on-1 until it's working, and if you're still not satisfied, we'll refund you in full. We're confident in what happens when you show up and build with us.

Justin Day

Founder, Day by Day (local SEO)

Runs a local-SEO agency for home services; built the Invisible Business Scan and an overnight 'Dreaming' system.

Darby Rollins

Chief Agent Officer, Agentic Society

Builds the Agentic OS — skills and agents for the community; founder of GenAI University.

Oscar Hidalgo

CEO, CG Service Pros · Vindex Consulting

Runs an Austin HVAC company and an AI consulting practice; built 'Ali,' the always-on outreach employee behind two $200K+ rebate-program contracts.

Sagun Sedai

Founder, Real Austin Report

Austin's one-man local news engine — an AI research-and-script pipeline behind millions of monthly views and a 66-week newsletter streak.

Alan VanToai

Founder, AI Performance Lab

Builds enterprise sales workflows: agents that research and grade accounts, then deliver AI assessments to executives.

Paul H. Graham

Founder, EG Capital

Built a CRM, media, and sales connector modeled after the Walt Disney strategy map.

Andrew Sponsler

AI & Automations Consultant

Builds story-driven lead generation agents with self-learning feedback loops from web analytics and engagement.

Peter Hewlett

Founder, PCH Conversions

Conversion specialist experimenting with multi-LLM research councils that debate their way to a better answer.

Chrissy McDannell

Founder, The Magnolia Firm

Founder of The Magnolia Firm and Aiic.Club — and owner of the NFC-tag grocery agent.

Ashley Kirchner

Founder, Decked Out Factory

Runs manufacturing and ecommerce companies on an agent-built ERP with hundreds of automations.

Doug Greenberg

Founder, PNW Advisory

Builds LocalFrequency.app (all Austin local events in one place) plus an AI reviewer and advisory dashboard.

Alexander Felice

Founder, Threadquarters

Built an internal company app for scheduling, time clocking, payroll, and Shopify-synced weekly product drops.

Mike Couvillion

Founder, Alexa Guru

30 years as a CTO across SaaS companies, now advising private-equity firms on AI rollouts and technical due diligence. Runs a fleet of Claude and Codex agents behind an orchestrator that assigns work to developer agents at the right level.

Jesse McElvain

Founder, Distru

Runs agents that review sales call transcripts for rep coaching.

David Boskovic

Founder, Obvious

Automating product development with AI agents.

Lucas Ferrer

Founder, Imagenyx

Builds agentic applications for HIPAA-compliant clinical workflows.

Wilson Smith

Runs live multi-round debates between Claude, Codex, and Gemini on every build — and audits every skill with a red-team agent. If he wouldn't put it on a Times Square billboard, it doesn't go in his LLM.

Brian Folsom

Founder, Inventory Sheets

Works on governance, provenance, and staged SOPs for agentic systems.

Damon Flowers

Modern Operations Expert, Modern Operators

Helping founders and teams scale faster with less stress. 4x CEO, 2x 8-figure exits.

Chance Forman

Luxury construction, Austin

Builds custom homes, wine cellars, and gyms in Austin; rolling AI across his teams to make people superhuman.

Mitchell Zoll

Attorney, Zoll Firm

Balancing fast-moving AI with the legal constraints lawyers face around privileged information.

Jo Feathers

Wellness for autoimmune disease

Reaches millions with authentic, lived-experience wellness content; co-owns a digital agency with Justin Day.

Cristian Aguilar

Agentic Society team

On the Agentic Society team.

Austin Distel

Founder, Agentic Society

Founder of the Agentic Society — a private room of owners and operators becoming Agentic CEOs.