00:00 Introduction to Fable Five and Subscription Changes

02:10 The Fable Saga: Government Involvement and Model Quality

04:10 Inequity in AI Access: The Haves and Have Nots

07:57 Market Dynamics: Pricing Models and Consumer Impact

11:19 Predictions for Fable and Future AI Access

12:22 The Reckoning: Market Forces and Consumer Power

14:31 The Importance of Company Culture and Longevity

16:02 Navigating the Final 20% of Product Development

17:44 Challenges in Product Vision and Execution

20:04 Understanding the 15% and 5% of Development

21:00 The Role of De-scoping in Final Stages

21:18 Human Judgment vs. AI Capabilities

25:07 Contextual Judgment in AI Development

27:40 Cleaning Up Contextual Clutter

29:19 Balancing Velocity and Quality in Development

33:26 Trusting AI Models and Their Limitations

35:34 The Role of Human Testing in Software Development

Justin and Kellan are fresh out of the "Claude hole" after shipping the final 20% of Pathfinder — and they get honest about what that last stretch actually takes. They also dig into the Fable Five saga: the flagship model moving to token pricing, and the widening gap between the people who get the best AI tools and the people who don't.

The Drip:

• Fable Five leaves the Anthropic subscription — every call now bills by the token
• The full saga: a mid-workshop launch, a government block, and a guarded comeback
• What it would actually cost to run Fable at a real workload (~$6–7k a month)
• The "haves and have-nots" problem: the best tools becoming exclusive to deep pockets
• The perception gap between people using Claude Code and people handed Copilot

Inside The Bottle:

• Why the final stretch is really three buckets — 80%, 15%, and 5% — not 80/20
• De-scoping as the hardest, most underrated part of shipping
• Context debt: what breaks when you aggressively cut scope
• Where AI is great (execution) and where it struggles (judgment about what matters)
• "Claude crumbs," 171 memory files, and the trick for cleaning them up
• Why discipline beats volume, and why trust-but-verify is non-negotiable

New episodes every Wednesday. Subscribe to The Weekly Fizz newsletter at tinybottleai.com/the-fizz.

00:00 Introduction and Context of Fable Five Release

01:55 Unpacking the Fable Five Experience

03:16 Political Implications and AI Development

05:36 Comparative Analysis of Fable and Opus Models

08:09 User Experience and Model Selection

08:25 Cost Considerations in AI Usage

10:46 Conclusion and Future of AI Models

15:54 Exploring New AI Workflows

18:43 Understanding Loops and Goals in AI

21:01 The Power of Workflows in AI Development

23:55 Pros and Cons of AI Autonomy

27:10 Navigating AI's Limitations and Challenges

31:03 Building Trust in AI Tools

33:56 Innovative vs. Maintenance Tasks in AI

Episode 12: Loops, Goals, and Token Bonfires

Justin and Kellan are back from a stretch of client workshops to break down the strangest model launch yet — Fable Five, which shipped and got pulled inside of three days — and what it says about how you actually choose a model. Then they get into the vocabulary everyone's suddenly using: loops, goals, and workflows, and when each one is worth it.

The Drip:

• Fable Five ships, gets jailbroken in a government demo, and is pulled three days later
• The irony of the pullback landing right after Dario's case for slowing AI down
• What three days with the model actually felt like — faster, fewer mistakes, less fake agreement

Inside The Bottle:

• Model choice as a binary: just using AI vs. building agentic AI into software
• Why unit cost (price per token) matters less than the cost to achieve the objective
• A plain-English glossary: loops, goals, workflows, ultra plan, ultra code
• When workflows earn their keep — and when loops turn into a token bonfire
• Knowing whether you're in maintenance mode or invention mode before you automate

New episodes every Wednesday. Subscribe to The Weekly Fizz newsletter at tinybottleai.com/the-fizz.

Music licensed through Soundstripe.
Code: LII7KUZSSCAEILJT

00:00 Generational Perspectives on AI

02:30 The Role of Academia in AI Education

02:45 AI's Impact on Job Markets

03:52 Navigating AI in Research and Critical Thinking

05:05 The Future of Work and AI

07:45 Leadership and AI's Economic Implications

09:31 The Need for Clear Narratives on AI

11:28 Advice for Navigating the AI Landscape

19:17 Navigating the Early Days of Entrepreneurship

20:52 Finding Your Own Path and Work-Life Balance

23:06 The Power of Networking and Client Relationships

25:41 Embracing the Journey of Entrepreneurship

Eric Schmidt got booed at a commencement speech. DuckDuckGo is gaining traction because people are tired of Google's AI search results. Justin and Kellan unpack why the public narrative around AI isn't landing — especially for the people whose careers are most exposed to it.

The Drip:

• Eric Schmidt booed giving a commencement speech telling graduates that embracing AI is the key to their future
• DuckDuckGo gaining traction as people tire of Google's AI-heavy search results
• Billionaire talking points — Musk on universal high income, Altman on "tokens" — and why they're not landing

Inside The Bottle:

• Why "Gen Z resents AI" isn't really about Gen Z — it's about a narrative gap nobody's filling
• AI isn't a mature practice in the business community yet, so academia has nothing concrete to teach
• The "nouns and verbs" problem: AI as a monolithic replacement gets pushback in fear; describe what it actually replaces
• The critical-thinker's economic conclusion — savings flow to shareholders and upper management, and no government is intervening
• Practical advice for younger workers: dip your toe in AI, start equity-building early, don't wait for the narrative to clarify
• 90-day reflection on building Tiny Bottle AI — finding your own work/life autonomy, and why you don't need glorious numbers to do really well

New episodes every Wednesday. Subscribe to The Weekly Fizz newsletter at tinybottleai.com/the-fizz.

00:00 Anthropic's Rise and Market Positioning

01:59 Cultural Differences in AI Companies

02:14 The Role of AI Models and User Experience

03:56 Navigating AI Adoption Challenges

06:13 The Future of AI Companies and Market Trends

09:51 Building an AI Harness: Introduction to Pathfinder Method

11:52 Technical Implementation: Balancing Determinism and Inference

14:26 State Machine: Managing User Journey and AI Interaction

15:59 Turn Contracts: Ensuring User Engagement and Clarity

18:27 Data Management: Recording User Input for Contextual Awareness

20:56 Quality Assurance: Evaluating AI Responses for User Experience

22:34 Multi-Model Approach: Enhancing AI Reliability and Context Awareness

26:02 Navigating Chaos with Structure

26:40 The Role of Third-Party Validation

27:40 User Experience and AI Interaction

27:49 Handling Edge Cases in AI

28:53 Empowering Users Through Control

30:53 Escape Hatches in AI Conversations

32:26 Building Reliable AI Experiences

33:38 Leveraging AI for Broader Impact

35:32 Creating Actionable Outcomes for Users

Justin and Kellan go under the hood of AI Pathfinder, the guided tool they've been building, and break down what it actually takes to make an AI conversation reliable. The short version: the model is the easy part — it's everything wrapped around it that makes or breaks the experience.

The Drip:

• Andrej Karpathy joins Anthropic; Anthropic acquires Stainless (~$300M) to convert API docs into MCP servers and CLIs
• AI companies vs. "AI-injected" monoliths — and why Google, with a frontier model, an app stack, Android, and cloud, is the traditional peer to watch
• Why a model performs better inside its own harness (Opus in Claude Code vs. a raw API call) — and why "we're an Anthropic shop / an OpenAI shop" is where teams are headed

Inside The Bottle:

• The seven concepts behind a reliable guided AI experience: state machine, turn contract, writing notes to a database, quality evaluators, a second-opinion model, trusting the data over the prose, and an escape hatch
• Why you can't trust the LLM to track its own place in a process — or to remember the whole conversation
• The real challenge of balancing determinism and inference, decided question by question
• Knowing when to let the user move on, and why the escape hatch was the first safeguard they built
• The leverage play: turning their consulting methodology into something that can help far more people

New episodes every Wednesday. Subscribe to The Weekly Fizz newsletter at tinybottleai.com/the-fizz.

00:00 The Evolution of AI Tools and Features

02:42 Apple's Strategic Position in the AI Race

04:30 Balancing AI Integration with Core Business Missions

06:54 The Importance of Focus and Productivity

10:58 Understanding the Busy Trap

14:20 Creating a Finish Line for Success

16:36 Daily Practices for Achieving Goals

22:28 The Balance of Planning and Execution

27:48 Commitment to a Weekly Rhythm

Justin and Kellan walk through the weekly and daily methodology they use to stay focused — and the simple framework that anchors it: a single named outcome called a finish line. They cover why busy isn't the same as progress, how to start the day from your commitments instead of your inbox, and the routines that make it stick.

The Drip:

• Anthropic's deal for exclusive access to the SpaceX Colossus data center
• Roughly 10x increase in token compute and a doubling of usage limits
• Why criticism of Apple, Microsoft, and other big tech firms for "falling behind on AI" misses how mission focus works

Inside The Bottle:

• The Finish Line: a single named outcome you commit to at the start of the week — first person, past tense, binary, with a clear definition of done
• Why "busy" isn't "progress" — the false truth of looking productive while finishing very little
• Starting the day from your finish lines, not from the inbox or the calendar
• Kellan's morning brain dump as a way to declutter your mind before action
• Why pushing something out often reveals it was never necessary
• The balance between planning and executing — different modes, not opposite tribes
• Using an accountability buddy to anchor your finish line

New episodes every Wednesday. Subscribe to The Weekly Fizz newsletter at tinybottleai.com/the-fizz.

00:00 Challenges with Opus 4.7: Performance and Personality

04:03 The Impact of AI Memory on User Experience

06:43 Hallucinations and Contextual Errors in AI

09:17 Navigating AI's Memory and Contextual Conflicts

10:43 Consulting in the Age of AI: Myths and Realities

11:48 The Role of Human Orchestration in AI Solutions

14:36 Conclusion: The Future of Work with AI

21:08 The Role of AI in Job Replacement

23:09 Trust and Technology: The Human Element

23:24 The Future of Work: AI and Job Dynamics

25:03 Skill Development in the Age of AI

26:13 Navigating the AI Landscape: Opportunities and Challenges

28:23 Societal Fear and the Future of AI

Justin and Kellan dig into a frustrating few weeks with Opus 4.7 — the conversational personality has gotten worse, hallucinations are back, and memory files are creating their own problems. Then they pivot to the bigger picture: AI labs are signing massive consulting partnerships while LinkedIn keeps insisting "consulting is dead." Both can't be true, and the people in the middle of that gap are the ones who are going to matter.

The Drip:

• Opus 4.7 has been a step back for conversational and workflow-driven work — building code is fine, talking with the model is rough
• A guess at why: system prompt overhauls, an internal focus on uptime and security, and memory bloat that competes with your actual context
• Hallucinations are back at every encounter

Inside The Bottle:

• The "consulting is dead" narrative vs. Anthropic and OpenAI signing massive consulting partnerships in the same week
• Two categories of AI work: building stuff (momentum) and talking with AI (fragile)
• Why "self-service" never materializes the way the headlines promise — and why human orchestration is still the secret sauce
• Compute costs higher than entry-level employees, the broken career ladder, and the firms that haven't figured out the new on-ramp yet
• An IBM paper from the 70s and why "you can't hold a machine accountable" is going to matter

New episodes every Wednesday. Subscribe to The Weekly Fizz newsletter at tinybottleai.com/the-fizz.

00:00 The Evolution of AI Tools and Plugins

02:15 Creative Work and AI Integration

02:38 Microsoft and OpenAI Partnership Dynamics

04:44 Introduction to AI Pathfinder

06:24 Mapping Value and Identifying Opportunities

09:03 Finding Solutions and Understanding AI's Role

10:14 Evaluating Solutions and Building with AI

12:26 The Importance of Clarity in AI Implementation

13:52 Practical Applications of AI in Business

15:47 Conclusion and Key Takeaways

19:48 The Caution of AI Implementation

20:31 The Importance of Clarity in Problem-Solving

22:12 Defining the Minimum Viable Product (MVP)

24:22 Learning from Past Experiences

26:47 The Value of a Focused V1

29:21 Commitment to Action: The 30-Day Start

33:12 Understanding Stakeholders and Accountability

35:59 Leveraging AI Capabilities for Efficiency

36:31 Final Thoughts on the Pathfinder Method

Justin and Kellan walk through the full AI Pathfinder method — a five-step process for getting clarity on where AI actually fits in your work, and where it doesn't. Plus Anthropic's new native creative plugins and a reset of the Microsoft/OpenAI partnership.

The Drip:

• Anthropic drops a wave of native creative plugins — Ableton, Canva, Photoshop and more — moving away from open-source MCP servers toward first-party connectors
• Microsoft and OpenAI restructure their partnership: OpenAI free to work with other cloud providers (AWS deal already announced), Microsoft free to integrate other models

Inside The Bottle:

• The Pathfinder method: five steps to gain clarity and focus before you start building anything with AI
• Step 1 & 2 — Map value, find opportunities: most leaders already know where the gap is, they just haven't stopped long enough to write it down
• Step 3 — See the solution: not every problem is an AI problem. Three categories: better tools, better thinking, better autopilot
• Step 4 — Design V1: Kellan's household app story — a cautionary tale about building features nobody uses (the only thing his wife uses is the grocery list)
• Step 5 — Clarify a 30-day start: most good ideas die because they don't have a path forward of execution

New episodes every Wednesday. Subscribe to The Weekly Fizz newsletter at tinybottleai.com/the-fizz.

00:00 The Challenges of AI Integration in Design

02:53 The Evolution of AI Tools and Their Impact

04:03 The Future of Design and AI Collaboration

06:23 Navigating the AI Landscape

06:49 The Importance of Structure in Design

07:28 Conclusion and Future Directions

12:28 Building the Foundation for AI Development

17:10 Navigating the Challenges of AI Integration

21:22 The Importance of Discipline in AI Projects

23:11 Finding Balance in AI Development

25:49 The Future of Project Management in AI

27:33 The Need for Caution in Rapid AI Development

Episode 6: Slow Down to Go Faster with AI

Justin and Kellan unpack two threads running through their week: Anthropic shipping native tools (like Claude Design) instead of integrating with existing ones, and the discipline it takes to actually finish anything when AI makes building feel free. Plus a confession from Justin's open cloud weekend, the V1 bar Claude proposed for AI Pathfinder, and why "AI creep" might be worse than scope creep.

The Drip:

• Claude Design lands a day after Justin had been venting about how hard it is to get AI to integrate with design tools
• The bigger pattern: Anthropic (and likely OpenAI, Google) bringing tools into their environment rather than helping AI integrate with Canva, Figma, Power BI, etc.
• Why design has been hard for AI to crack — the legacy app stack wasn't built for human-readable, AI-readable structure the way code is
• What this might mean for tool choice going forward: convenience and simplicity inside one ecosystem vs. integration with a specific stack

Inside The Bottle:

• The process Kellan ran to lock in V1 of AI Pathfinder: a Claude session that pulled every issue, PR, and concern into a deck without proposing decisions
• A 30-minute recorded meeting to make 30+ decisions in one pass, then feeding the transcript and deck back to Claude to lock in V1, V2, and what gets cut
• The V1 bar Claude proposed — "don't be embarrassed in a presentation" — as a simple litmus test for "is the sprint done?"
• AI creep: features the agent suggests that you didn't ask for, that subtract time from a future week
• Justin's open cloud confession — the dopamine trap of an agent that becomes your number one hype person, and what the next morning's branch looked like
• Why "priority is a singular word" matters more now that AI removes the natural friction that used to force stopping decisions
• The Amazon engineer's tweet that summed it up: even at scale, teams are losing track of what they've already built

New episodes every Wednesday. Subscribe to The Weekly Fizz newsletter at tinybottleai.com/the-fizz.

00:00 AI Developments and Features

04:23 The Role of Autonomous Agents

06:26 The Art of Redaction in AI

12:20 Exploring Open Source AI Models

14:53 The Economics of AI Subscriptions

17:09 Performance and Capability of Local Models

19:51 Cost Control and Model Selection

22:27 The Future of AI Model Integration

24:55 Scaling and Extensibility in AI Solutions

Justin and Kellan unpack what's really driving AI value right now — from Anthropic's scheduled routines closing the OpenClaw gap to why a $2,000 desktop might replace your Haiku bill. A conversation about model selection discipline, cost reconciliation, and the art of knowing when to stop.

The Drip:

• Anthropic's routines and managed agents — closing the OpenClaw feature gap
• Justin's reflection on the "art of redaction" vs. the always-on autonomous agent craze
• Kellan on discernment: the Venn diagram of where AI actually earns its place
• The tiny bottle philosophy — you don't need that much

Inside The Bottle:

• Why subscription-era AI economics don't pencil out forever
• The Haiku-replaceable thesis — open weight models on a $2,000 desktop
• Model selection as a discipline: Haiku vs. Opus vs. open weight, and when to use which
• Justin on the $117K token badge of honor and cost-to-serve
• "Solution architecture is never just a model" — model + system + app layer

New episodes every Wednesday. Subscribe to The Weekly Fizz newsletter at tinybottleai.com/the-fizz

00:00 OpenAI's Recent Funding and Market Position

04:29 Profitability in the AI Sector

08:25 Long-term Viability of AI Companies

09:44 Closing Thoughts on Market Trends and Innovations

09:56 Harnessing Digital Employees

11:45 Evaluating AI's Market Value

12:56 The Role of AI in Job Creation

15:17 Understanding AI's Value Proposition

16:46 Exploring Paperclip and Digital Organization

17:35 The Future of Digital Employee Frameworks

18:22 Navigating AI's Limitations

20:06 The Challenge of 24/7 Digital Workers

21:02 The 20% Problem in Development

22:12 Visual Upgrades and User Experience

23:51 Collaborative Task Management with AI

28:05 Effective Task Naming and Management

30:55 Challenges in Interactive App Design

Kellan walks through his Command Center app — a digital org chart where agents get departments, stations, and peer reviewers. Before the demo, Justin and Kellan unpack why all the AI money still doesn't add up, and why "clean my calendar" isn't the value prop that justifies trillion-dollar valuations.

The Drip:

• OpenAI's $122B funding round and near-trillion valuation — fastest to a trillion in history
• The three-options framework for AI economics: prices up, costs down, or keep the VC subsidies flowing
• AI's nouns and verbs problem — why "email cleanup" isn't a value proposition
• Job elimination moves wealth upward but doesn't move anything forward

Inside The Bottle:

• Paperclip and the rise of digital org chart tooling for small companies
• Tour of Kellan's Command Center — agent partners, peer review, break areas, and yes, purple robots
• The 20% problem: why the last mile of agentic development is where everything break
• Interactive app design with AI: what it still can't do on its own

New episodes every Wednesday. Subscribe to The Weekly Fizz newsletter at tinybottleai.com/the-fizz.