Podcast Summary: How I AI

Episode: How to build your own AI developer tools with Claude Code | CJ Hess (Tenex)
Date: February 9, 2026
Host: Claire Vo
Guest: CJ Hess (Tenex)

Overview

In this episode, host Claire Vo sits down with CJ Hess from Tenex to explore building personal AI developer tools using Claude Code. CJ demonstrates his hands-on approach to crafting developer workflows and tools—especially his app "Flowy"—around AI, revealing how to deeply integrate models like Claude into daily engineering tasks. The episode is technical yet accessible, focusing on practical steps to streamline development, increase productivity, and develop custom workflows for modern AI-powered coding. Throughout, both share tips, workflows, and candid commentary on the state and future of AI engineering.

Key Discussion Points & Insights

Why Claude Code?

Delight and Steerability: CJ explains his preference for Claude Code due to its steerability and strong intent understanding, which he feels surpasses alternatives like GPT-5.2 for "real software engineering" tasks (03:32–04:47).

“Working with Claude is just such a delight… it just feels so steerable. And… when I want to dig deep, it just does it.” — CJ Hess (03:32)

Personalization of Dev Environments

Hyper-Custom Workflow: Modern AI tools let developers build hyper-personalized workflows; pre-AI, choices were limited to standardized IDEs and linters. Today, devs can assemble their own AI tool ecosystems for productivity (04:47–05:29).

“You could have a totally different AI engineering workflow than your colleague… It’s making you individually a lot more efficient and effective.” — Claire Vo (04:47)
AI for “Chore Problems”: Tasks like fixing linter configs are no longer “forever problems”; Claude can quickly resolve formerly tedious set-up and environment issues (05:29–06:14).

Underappreciated Use Case: Dev Environment Setup

CJ and Claire discuss using Claude Code to automate onboarding, environment set-up, and repo understanding—even for non-technical team members (06:14–07:06).

“Just ask Claude code to do it. Say like help me understand this repo and get my computer set up to run…” — Claire Vo (06:14)

Planning & Visual Thinking

Evolution from Markdown to Visualizations:
- The classic plan is iteratively crafted in markdown, but CJ struggled with the limitations of ASCII diagrams and markdown for complex flows.
- CJ built “Flowy,” a tool where flowcharts/UI mockups can be defined via JSON, removing reliance on brittle ASCII or Mermaid syntax (07:06–10:44).
“I really like this visual way to think about things, but I really hate staring at these ASCII diagrams… So I’ve played around with this tool to basically give Claude these JSON files.” — CJ Hess (07:40)
Flowy in Practice:
- Flowy generates UI mockups and flowcharts dynamically through Claude model prompts and a custom JSON schema.
- Acts as both editor and visualization tool—edits are saved as JSON, easily referenced and updated by Claude for rapid iteration (10:44–11:55).
“You can then point Claude back at it and say, ‘Hey, I know you did this, but actually let’s say I want a step here…’”—CJ Hess (11:26)

Building Claude “Skills” for Custom Tools

Custom “skills” (markdown docs) tell the model how to handle proprietary formats (e.g., Flowy JSON). CJ iteratively refines these as he runs into edge cases or issues (12:17–14:10).

“And I find that works better than something like Mermaid… because I really feel the power of building my own dev tools now—and that I really don’t want to hit the constraints of Mermaid…” — CJ Hess (13:35)

The DIY Trend in Dev Tools

“Build vs Buy” Flips:
- With AI, building custom tools is so easy and cheap that it’s often preferable to buying SaaS (14:10–15:26).
“It’s almost not worth spending the extra money anymore…everyone’s posting some product, some ridiculous pricing tier and saying ‘someone please vibe code this.’” — CJ Hess (15:08)

Live Walkthrough: Using Flowy with Claude Code

Workflow Demo (15:32–36:51)

CJ’s Process:
1. Prompting the Model: Uses aliases in the terminal (e.g., “Kevin” for bypass permissions) to interact with Claude, providing high-level instruction and context refresh (16:10–17:20).
2. Generating Flowcharts: Prompts Claude—using skills—to create both animation sequences and user flows (17:20–25:05).
3. Editing and Iteration: Tweaks details (e.g., changing animation duration) directly in the Flowy UI and re-prompts Claude to update charts accordingly (25:05–27:28).
4. UI Mockups: Prompts for initial UI mockups, then iterates visually. Points out the ease of moving between human-friendly visual form and model-friendly code/markdown (31:44–33:14).
  
  “It’s almost like I want to see it visually and Claude wants to see it as markdown so we can kind of speak in our own way…” — CJ Hess (30:26)
Bypassing Planning: Increasingly skips explicit “plan” steps and asks Claude to build features directly from diagrams and mockups (“just build it”) (33:30–34:25).

“I’m going to skip the plan and say, based on the flowcharts and the mockups, build this feature…” — CJ Hess (33:30)
Iterative Review: Once code is generated, tests for functionality, and iterates as needed (34:25–36:51).

Takeaways

Rapid Prototyping: These workflows enable building useful features—complete with visual flow, UI, and logic—in a matter of minutes.
Flexibility: The system strongly supports rapid change, quick refinements via the interface or prompt, and easy interaction between visual and code states.

Model-to-Model Review: Claude + Codex

Quality Checks: CJ uses GPT Codex (aliased as “Carl”) to review Claude code, checking for discrepancies, best practices, or refactoring suggestions (36:51–44:01).

“At this point, I feel like I’m mostly a QA person, and if there’s something that’s logically wrong… Codex always finds those types of things.” — CJ Hess (37:07)
Prompts for Review:
1. “Does the code reflect the diagrams?”
2. “Any general code smells?”
3. “What would you refactor for clarity/extensibility?”
Twin-Model Approach: Codex serves as a critical “staff engineer,” offering fresh and sometimes tough feedback for further iteration (40:52–41:48).

“Codex is like kind of a really good curmudgeonly staff engineer that will look at your code and tell you what’s wrong with it…” — Claire Vo (41:12)

Broader Reflections

Visual Paradigms in AI Dev Tools

Recognizes a rising paradigm in AI tool use: mixed visual (human) and coded (AI) modalities for communication and iteration (30:26–31:44).

The Future: On-Demand Apps and Custom AI UIs

Imagines a world where AI agents generate and dispose of apps/interfaces on demand, supporting better collaboration and visualization between human and AI (31:02–31:44).

Blueprint for AI-Native Dev Workflows

Advocates for codifying communication—via skills and diagrams—to create a shared language and rhythm between human and AI agents for building and iterating rapidly and cleanly (36:32–36:51).

Notable Quotes & Memorable Moments

On the power of building your own dev tools:

“This is a dev tool that was almost 100% prompted.” — CJ Hess (09:09)
On shifting expectations:

“Yeah, shifting expectations on these models.” — CJ Hess (10:24)
On developer autonomy:

“Of course we should build this. Until we hit some constraint, we should build it…” — Claire Vo (14:10)
On AI model feedback personalities:

“With the Google models, I also always used to say is they were like very smart, but clinically depressed…” — Claire Vo (41:32)

Lightning Round

(45:29–52:36)

Most Excited About in AI:
- CJ loves Google's Genie 3: a text-to-explorable-world generator, forecasting its potential to go viral with broader access (45:29–47:14).
Prompting Style When AI “Misbehaves”:

“I used to be a yeller… but I started to feel bad about it. So I’m like, ‘Good try. You did your best… here’s what I was going for… this is on me.’” — CJ Hess (50:01)
Can Others Use Flowy?
- CJ is working to release a version of Flowy for public use, likely with associated Claude skills (51:25).

Important Timestamps

03:32 – Why Claude/intent understanding
07:06 – Planning workflows, ASCII vs. Flowy
12:17 – Custom skills for proprietary tools
15:32 – Demoing Flowy live
25:05 – Editing diagrams in Flowy, handoff to Claude
31:02 – Future of visual human-AI collaboration
36:32 – Recap: Flowy workflow
36:51 – Using model-to-model review (Claude + Codex)
45:29 – Lightning round: excited about Genie
50:01 – Prompting techniques when models fail
51:25 – Plans for releasing Flowy to the public

Final Thoughts

This episode exemplifies the future of developer tooling: deeply custom, highly interactive, and symbiotic with AI agents. CJ’s approach—using Claude Code to generate plans, visuals, and production code while augmenting the process with custom “skills”—shows how engineers can embrace AI as a creative, collaborative partner and even build tools the AI itself helps maintain and extend. Both host and guest encourage listeners to take a hands-on, experimental approach: try building your own, iterate constantly, and use the feedback loop between humans and AI as both a productivity engine and source of delight.

Guest Links:

Twitter: @SEJAYHess
LinkedIn: SE JAY Hess

Host/Podcast:

howiaipod.com

(Fluffy outro and promotional segments excluded as per guidelines).

Podcast Summary: How I AI

Episode: How to build your own AI developer tools with Claude Code | CJ Hess (Tenex)
Date: February 9, 2026
Host: Claire Vo
Guest: CJ Hess (Tenex)

Overview

Key Discussion Points & Insights

Why Claude Code?

Delight and Steerability: CJ explains his preference for Claude Code due to its steerability and strong intent understanding, which he feels surpasses alternatives like GPT-5.2 for "real software engineering" tasks (03:32–04:47).

“Working with Claude is just such a delight… it just feels so steerable. And… when I want to dig deep, it just does it.” — CJ Hess (03:32)

Personalization of Dev Environments

Hyper-Custom Workflow: Modern AI tools let developers build hyper-personalized workflows; pre-AI, choices were limited to standardized IDEs and linters. Today, devs can assemble their own AI tool ecosystems for productivity (04:47–05:29).

“You could have a totally different AI engineering workflow than your colleague… It’s making you individually a lot more efficient and effective.” — Claire Vo (04:47)
AI for “Chore Problems”: Tasks like fixing linter configs are no longer “forever problems”; Claude can quickly resolve formerly tedious set-up and environment issues (05:29–06:14).

Underappreciated Use Case: Dev Environment Setup

CJ and Claire discuss using Claude Code to automate onboarding, environment set-up, and repo understanding—even for non-technical team members (06:14–07:06).

“Just ask Claude code to do it. Say like help me understand this repo and get my computer set up to run…” — Claire Vo (06:14)

Planning & Visual Thinking

Evolution from Markdown to Visualizations:
- The classic plan is iteratively crafted in markdown, but CJ struggled with the limitations of ASCII diagrams and markdown for complex flows.
- CJ built “Flowy,” a tool where flowcharts/UI mockups can be defined via JSON, removing reliance on brittle ASCII or Mermaid syntax (07:06–10:44).
“I really like this visual way to think about things, but I really hate staring at these ASCII diagrams… So I’ve played around with this tool to basically give Claude these JSON files.” — CJ Hess (07:40)
Flowy in Practice:
- Flowy generates UI mockups and flowcharts dynamically through Claude model prompts and a custom JSON schema.
- Acts as both editor and visualization tool—edits are saved as JSON, easily referenced and updated by Claude for rapid iteration (10:44–11:55).
“You can then point Claude back at it and say, ‘Hey, I know you did this, but actually let’s say I want a step here…’”—CJ Hess (11:26)

Building Claude “Skills” for Custom Tools

Custom “skills” (markdown docs) tell the model how to handle proprietary formats (e.g., Flowy JSON). CJ iteratively refines these as he runs into edge cases or issues (12:17–14:10).

“And I find that works better than something like Mermaid… because I really feel the power of building my own dev tools now—and that I really don’t want to hit the constraints of Mermaid…” — CJ Hess (13:35)

The DIY Trend in Dev Tools

“Build vs Buy” Flips:
- With AI, building custom tools is so easy and cheap that it’s often preferable to buying SaaS (14:10–15:26).
“It’s almost not worth spending the extra money anymore…everyone’s posting some product, some ridiculous pricing tier and saying ‘someone please vibe code this.’” — CJ Hess (15:08)

Live Walkthrough: Using Flowy with Claude Code

Workflow Demo (15:32–36:51)

CJ’s Process:
1. Prompting the Model: Uses aliases in the terminal (e.g., “Kevin” for bypass permissions) to interact with Claude, providing high-level instruction and context refresh (16:10–17:20).
2. Generating Flowcharts: Prompts Claude—using skills—to create both animation sequences and user flows (17:20–25:05).
3. Editing and Iteration: Tweaks details (e.g., changing animation duration) directly in the Flowy UI and re-prompts Claude to update charts accordingly (25:05–27:28).
4. UI Mockups: Prompts for initial UI mockups, then iterates visually. Points out the ease of moving between human-friendly visual form and model-friendly code/markdown (31:44–33:14).
  
  “It’s almost like I want to see it visually and Claude wants to see it as markdown so we can kind of speak in our own way…” — CJ Hess (30:26)
Bypassing Planning: Increasingly skips explicit “plan” steps and asks Claude to build features directly from diagrams and mockups (“just build it”) (33:30–34:25).

“I’m going to skip the plan and say, based on the flowcharts and the mockups, build this feature…” — CJ Hess (33:30)
Iterative Review: Once code is generated, tests for functionality, and iterates as needed (34:25–36:51).

Takeaways

Rapid Prototyping: These workflows enable building useful features—complete with visual flow, UI, and logic—in a matter of minutes.
Flexibility: The system strongly supports rapid change, quick refinements via the interface or prompt, and easy interaction between visual and code states.

Model-to-Model Review: Claude + Codex

Quality Checks: CJ uses GPT Codex (aliased as “Carl”) to review Claude code, checking for discrepancies, best practices, or refactoring suggestions (36:51–44:01).

“At this point, I feel like I’m mostly a QA person, and if there’s something that’s logically wrong… Codex always finds those types of things.” — CJ Hess (37:07)
Prompts for Review:
1. “Does the code reflect the diagrams?”
2. “Any general code smells?”
3. “What would you refactor for clarity/extensibility?”
Twin-Model Approach: Codex serves as a critical “staff engineer,” offering fresh and sometimes tough feedback for further iteration (40:52–41:48).

“Codex is like kind of a really good curmudgeonly staff engineer that will look at your code and tell you what’s wrong with it…” — Claire Vo (41:12)

Broader Reflections

Visual Paradigms in AI Dev Tools

Recognizes a rising paradigm in AI tool use: mixed visual (human) and coded (AI) modalities for communication and iteration (30:26–31:44).

The Future: On-Demand Apps and Custom AI UIs

Imagines a world where AI agents generate and dispose of apps/interfaces on demand, supporting better collaboration and visualization between human and AI (31:02–31:44).

Blueprint for AI-Native Dev Workflows

Advocates for codifying communication—via skills and diagrams—to create a shared language and rhythm between human and AI agents for building and iterating rapidly and cleanly (36:32–36:51).

Notable Quotes & Memorable Moments

On the power of building your own dev tools:

“This is a dev tool that was almost 100% prompted.” — CJ Hess (09:09)
On shifting expectations:

“Yeah, shifting expectations on these models.” — CJ Hess (10:24)
On developer autonomy:

“Of course we should build this. Until we hit some constraint, we should build it…” — Claire Vo (14:10)
On AI model feedback personalities:

“With the Google models, I also always used to say is they were like very smart, but clinically depressed…” — Claire Vo (41:32)

Lightning Round

(45:29–52:36)

Most Excited About in AI:
- CJ loves Google's Genie 3: a text-to-explorable-world generator, forecasting its potential to go viral with broader access (45:29–47:14).
Prompting Style When AI “Misbehaves”:

“I used to be a yeller… but I started to feel bad about it. So I’m like, ‘Good try. You did your best… here’s what I was going for… this is on me.’” — CJ Hess (50:01)
Can Others Use Flowy?
- CJ is working to release a version of Flowy for public use, likely with associated Claude skills (51:25).

Important Timestamps

03:32 – Why Claude/intent understanding
07:06 – Planning workflows, ASCII vs. Flowy
12:17 – Custom skills for proprietary tools
15:32 – Demoing Flowy live
25:05 – Editing diagrams in Flowy, handoff to Claude
31:02 – Future of visual human-AI collaboration
36:32 – Recap: Flowy workflow
36:51 – Using model-to-model review (Claude + Codex)
45:29 – Lightning round: excited about Genie
50:01 – Prompting techniques when models fail
51:25 – Plans for releasing Flowy to the public

Final Thoughts

Guest Links:

Twitter: @SEJAYHess
LinkedIn: SE JAY Hess

Host/Podcast:

howiaipod.com

(Fluffy outro and promotional segments excluded as per guidelines).

How to build your own AI developer tools with Claude Code | CJ Hess (Tenex)

Powered by Wave AI

Summary

Podcast Summary: How I AI

Overview

Key Discussion Points & Insights

Why Claude Code?

Personalization of Dev Environments

Underappreciated Use Case: Dev Environment Setup

Planning & Visual Thinking

Building Claude “Skills” for Custom Tools

The DIY Trend in Dev Tools

Live Walkthrough: Using Flowy with Claude Code

Workflow Demo (15:32–36:51)

Takeaways

Model-to-Model Review: Claude + Codex

Broader Reflections

Visual Paradigms in AI Dev Tools

The Future: On-Demand Apps and Custom AI UIs

Blueprint for AI-Native Dev Workflows

Notable Quotes & Memorable Moments

Lightning Round

Important Timestamps

Final Thoughts

Summary

Podcast Summary: How I AI

Overview

Key Discussion Points & Insights

Why Claude Code?

Personalization of Dev Environments

Underappreciated Use Case: Dev Environment Setup

Planning & Visual Thinking

Building Claude “Skills” for Custom Tools

The DIY Trend in Dev Tools

Live Walkthrough: Using Flowy with Claude Code

Workflow Demo (15:32–36:51)

Takeaways

Model-to-Model Review: Claude + Codex

Broader Reflections

Visual Paradigms in AI Dev Tools

The Future: On-Demand Apps and Custom AI UIs

Blueprint for AI-Native Dev Workflows

Notable Quotes & Memorable Moments

Lightning Round

Important Timestamps

Final Thoughts