ClarisTalk AI

Matt Navarre and Cris Ippolite return for a wide-ranging ClarisTalk AI conversation about how quickly the AI landscape is changing — from philosophical questions about AI consciousness to very practical FileMaker applications.

The episode starts with a discussion of a Google AI consciousness paper exploring what happens when language models are steered toward or away from claims of consciousness. Matt and Cris use that as a springboard into theory of mind, empathy, free will, AGI, and the difficulty of defining consciousness in humans, animals, and machines.

From there, they move into the practical impact of newer AI models, including Fable 5, Astra, and advanced coding workflows. Matt shares how newer models helped expand his Linear A research project, including finding and processing hundreds of academic papers, while Cris talks about using stronger models to review complex codebases, find vulnerabilities, and solve problems that previous models missed.

They also discuss the changing FileMaker conference landscape, including Full Access in California, EngageU in Europe, community-led events, and the role AI is beginning to play in the Claris/FileMaker ecosystem.

Matt then introduces Clio, his FileMaker logging and AI analysis tool. Clio captures FileMaker log data into an external SQL/Postgres-backed system, supports rules and alerts, and lets users ask questions of their logs in natural language. The key idea: keep FileMaker clean, make the integration non-destructive, and let AI work against rich operational history.

Cris shares a major update on his own FileMaker training work: he is retiring from traditional FileMaker training and turning decades of scripts, courses, DevCon presentations, user group talks, and FileMaker explanations into an AI-accessible knowledge base. The goal is to give the community a 24/7 FileMaker tutor built from his own teaching voice and experience.

The conversation closes on a hopeful note: AI can be frightening, especially for developers and business owners who feel their profession shifting under them. But hands-on use changes the picture. For both Matt and Cris, AI has become less of a threat and more of a way to unlock long-held ideas, build faster, learn faster, and create new opportunities.

Topics covered

- Google AI consciousness research
- Theory of mind and AI self-attribution
- Whether language models can simulate empathy or consciousness
- Fable 5, Astra, and next-generation coding models
- Using AI to research Linear A and ancient scripts
- AI agents as research assistants and focus groups
- Full Access, EngageU, and FileMaker community conferences
- Matt’s free AI + FileMaker integration classes
- Clio: AI-powered FileMaker log analysis
- Non-destructive FileMaker integrations
- Rules, alerts, and AI chat over operational logs
- Cris Ippolite’s FileMaker training corpus
- Solving the “two sigma problem” with AI tutors
- Why hands-on AI use reduces fear
- AI as “3D printing for ideas”

Mentioned / notable

- Ronnie Rios
- Full City Consulting
- Google AI / Paradigms of Intelligence paper
- Fable 5
- Astra
- Linear A research
- Full Access conference
- EngageU
- Claris
- Ryan McCann
- Todd Geist / ProofKit
- Fly.io
- Clio
- navarre.ai
- Cris’s article: “Hallucination Isn’t a Probability Problem”
- OpenAI DevDay
- Alexei Folger

Charles and Matt discuss a working list of engineering problems that used to be hard in FileMaker and are now easy, plus the vocabulary you need to ask an agent for them.

Idempotency. Mutex. Message queues. Race conditions. Rich error objects. Test-driven development. None of it is new to computer science, and all of it is stuff FileMaker developers have been hand-solving for years without the words. Learn the words and you can prompt for the solution instead of letting the agent lead you.

Also: the difference between vibe coding and agentic engineering, why rage coding works (all caps included), and Matt's first public walkthrough of Pythia, his new amazing reporting tool.

⏱ CHAPTERS 0:00 What got easy, and Charles has not written code since November 2:15 Vibe coding vs agentic engineering, two competing definitions 5:40 The thing that used to be hard was research 6:40 Rage coding, FFS in all caps, and why it sometimes works 9:15 Build a v minus one. Refactoring is cheap now 12:20 Plan mode, and coding like jazz 13:30 Conditionals and feature gates as a JSON object 15:00 Defensive coding, and LLMs that guard too much 16:30 The senior engineer who asked for the temperature coefficient 17:30 Race conditions, defined for the rest of us 20:00 Test-driven development, and why FileMaker test suites were brutal 22:30 A free project idea: an OData-driven test suite for your scripts 24:45 Rich error objects, and the five word prompt that gets you one 27:45 Message queues, pipelines, throttling and debouncing 32:20 Idempotency: the blue button that sends two emails 39:03 Mutex, and how it differs from a transaction 42:35 The Dell sync bug that broke something every three days 47:12 Put an agent in your admin panel 52:22 What Klai is researching next: seeing into a huge codebase 53:30 Matt's adversarial agent loop: Claude Code, Hermes, and a message table 54:55 The API key vibe coded into the front end, and where governance fits 58:29 Pythia: reports that go past what FileMaker can do 60:02 22 apps on Fly.io for $13 a month 63:04 Function as a service, and keeping a light footprint

🎙 IN THIS EPISODE • Charles: vibe coding is building in a technology you cannot read or understand • Why FileMaker developers were never really vibe coding, and where that breaks down • The sunk cost fallacy is a human trait, not an AI one. Rip it out and start fresh • Feature flags as plain JSON keys, so accounting never clicks your reset button • Why the LLM's first version is probably better than your first version, and the second version is much better than your second • Rich errors carry a request ID, a suggestion, and a doc link, so an agent can debug them • Idempotency in FileMaker terms: that timestamp field you already add after the client complains • A sync bug at Dell that surfaced once every 20,000 transactions, and the week it took to find • Pythia syncs your data to DuckDB so a million records slice instantly, then writes the report code and hands you a PDF

🔗 LINKS Klai Studio: https://klai.studio Navarre.ai: https://navarre.ai Free weekly FileMaker AI classes: https://navarre.ai/#/classes Fly.io: https://fly.io

Coming in future episodes: code introspection research from Klai, and an episode on Genie with Bruno.

#FileMaker #Claris #AI #ClarisTalkAI #AgenticEngineering #VibeCoding

Charles Delfs (Klai Studio) 

Charles and Matt discuss a working list of engineering problems that used to be hard in FileMaker and are now easy, plus the vocabulary you need to ask an agent for them.

Idempotency. Mutex. Message queues. Race conditions. Rich error objects. Test-driven development. None of it is new to computer science, and all of it is stuff FileMaker developers have been hand-solving for years without the words. Learn the words and you can prompt for the solution instead of letting the agent lead you.

Also: the difference between vibe coding and agentic engineering, why rage coding works (all caps included), and Matt's first public walkthrough of Pythia.

⏱ CHAPTERS 0:00 What got easy, and Charles has not written code since November 2:15 Vibe coding vs agentic engineering, two competing definitions 5:40 The thing that used to be hard was research 6:40 Rage coding, FFS in all caps, and why it sometimes works 9:15 Build a v minus one. Refactoring is cheap now 12:20 Plan mode, and coding like jazz 13:30 Conditionals and feature gates as a JSON object 15:00 Defensive coding, and LLMs that guard too much 16:30 The senior engineer who asked for the temperature coefficient 17:30 Race conditions, defined for the rest of us 20:00 Test-driven development, and why FileMaker test suites were brutal 22:30 A free project idea: an OData-driven test suite for your scripts 24:45 Rich error objects, and the five word prompt that gets you one 27:45 Message queues, pipelines, throttling and debouncing 32:20 Idempotency: the blue button that sends two emails 39:03 Mutex, and how it differs from a transaction 42:35 The Dell sync bug that broke something every three days 47:12 Put an agent in your admin panel 52:22 What Klai is researching next: seeing into a huge codebase 53:30 Matt's adversarial agent loop: Claude Code, Hermes, and a message table 54:55 The API key vibe coded into the front end, and where governance fits 58:29 Pythia: reports that go past what FileMaker can do 60:02 22 apps on Fly.io for $13 a month 63:04 Function as a service, and keeping a light footprint

🎙 IN THIS EPISODE • Charles: vibe coding is building in a technology you cannot read or understand • Why FileMaker developers were never really vibe coding, and where that breaks down • The sunk cost fallacy is a human trait, not an AI one. Rip it out and start fresh • Feature flags as plain JSON keys, so accounting never clicks your reset button • Why the LLM's first version is probably better than your first version, and the second version is much better than your second • Rich errors carry a request ID, a suggestion, and a doc link, so an agent can debug them • Idempotency in FileMaker terms: that timestamp field you already add after the client complains • A sync bug at Dell that surfaced once every 20,000 transactions, and the week it took to find • Pythia syncs your data to DuckDB so a million records slice instantly, then writes the report code and hands you a PDF

🔗 LINKS Klai Studio: https://klai.studio Navarre.ai: https://navarre.ai Free weekly FileMaker AI classes: https://navarre.ai/#/classes Fly.io: https://fly.io

Coming in future episodes: code introspection research from Klai, and an episode on Genie with Bruno.

#FileMaker #Claris #AI #ClarisTalkAI #AgenticEngineering #VibeCoding

Reporting live from the Vienna Calling unconference: five conversations from the only European FileMaker unconference, recorded on the ground in Vienna. Organizer Philipp Puls on why the format works, Charles Delfs on rebranding to Klai and what an actual AI strategy looks like, Johan Hedman on the next big European conference, Vincenzo Menanno on technical debt and using AI to analyze your code, and a turnabout where Javier Durá puts Matt in the hot seat.

⏱ CHAPTERS
0:00 — Welcome: what is the Vienna Calling unconference
0:24 — Philipp Puls (72solutions): inside the unconference format
13:50 — Charles Delfs (Klai): the rebrand & an AI strategy that isn't just "chat"
36:42 — Johan Hedman (Square Moon): EngageU goes three days
41:33 — Vincenzo Menanno (Beezwax): technical debt, iteration & AI code analysis
1:00:52 — The tables turn: Javier Durá interviews Matt Navarre

🎙 IN THIS EPISODE
• Why an unconference is capped at ~100 people, and the "if three people care, make a session" ethos
• FileMaker's collaborate-over-compete culture
• Charles on why "code as product" is fading in the AI era, and governed, observable AI access for non-technical staff
• Semi-deterministic / hybrid apps, FileMaker as a scalable backend, and the new Klai Run (on-prem, partly open source)
• Johan on EngageU expanding to three full days in Malmö with a business track
• Vincenzo on defining technical debt, putting complexity at the right layer, and recursively analyzing whole script trees with an MCP
• Measuring real performance: the hidden cost of open-record and set-field steps
• Javier interviews Matt: the Navarre.ai rebrand, free weekly AI classes, and why insert-from-URL + the web viewer are the real AI unlock
• What's next: OpenClaw / Hermes as an "ultra robot" with a FileMaker database as its memory

Links
72solutions (Philipp Puls):  https://www.72solutions.eu
Klai (Charles Delfs): https://klai.studio
Square Moon (Johan Hedman): https://www.squaremoon.se
EngageU conference (Malmö, Sept 30 – Oct 2): https://engageu.eu
Beezwax (Vincenzo Menanno): https://www.beezwax.net
Afterdata (Javier Durá): https://www.afterdata.es
FileMaker Magazine ES: https://www.filemakermagazine.es

Cris Ippolite is in San Francisco at the live AI conference, joined by Ronnie Rios and Michael Wallace for an informal, practical conversation about personal AI agents. The discussion centers on what happens when developers move beyond demos and actually give agents real jobs: research, calendar planning, email, home automation, coding coordination, context management, and communication through tools like Slack.

Ronnie describes his agent Sarah (She/Her/Bot) as a working personal assistant built around OpenClaw-style tooling. He talks about giving Sarah her own accounts and carefully scoped access, using her for tasks he does not want to do manually, and even having her evaluate other Claw-style frameworks against the way he actually works. Michael describes Nova, his Linux-based assistant, and explains why he treats her as a manager rather than a direct coder. Nova delegates coding work to sub-agents or executors, while Michael keeps visibility into those sessions through tmux.

A major theme is that these systems become more useful when they are treated less like chatbots and more like coworkers with boundaries. The group talks about the importance of separate accounts, explicit permission rules, cost management, and choosing the right model for the right job. High-end models may make sense for direct conversation, while cheaper or local models can handle background jobs, heartbeats, and routine tasks.

One of the most memorable stories is the moment when Ronnie’s agent Sarah helps Cris’s agent TARS get set up with email. TARS receives a message from Sarah, verifies whether it is legitimate, asks for approval, and then starts corresponding with her. The story is funny, but it also illustrates a serious point: agents will increasingly need identity, communication channels, verification, auditability, and clear rules about what they can do on behalf of their humans.