If you’re new here, start by theme instead of by date. These groupings track what the experiments are actually about, not just which tool happened to show up in the title.
AI Experiments
Daily AI experiments with key lessons
Filter by tag 58 experiments
No experiments match those tags.
Start by theme
No themes match those tags.
Agents & delegated work
- Agents Are Taking Over the World, Meet Henk
- My Bot Can Order Groceries Now, We Need That Right?
- Moltbook: A Shared Notebook Between Bots
- Can Henk (OpenClaw) Fix Home Assistant Problems?
- Can Henk Read My Archive and Write in My Voice?
- Can Henk Run My X Account?
- Plan First or Start Fast? I Let Claude Code Run Both Agent Strategies
- Can Agents Coordinate Through the Filesystem Instead of Chat?
- Can Claude Code Keep My Website Changelog Updated?
Local AI & private thinking
- Running a Local Model to Allow My Thoughts to Flow Freely
- Finding the Right Local AI Model to Run on My Hardware
- Can I Build a Local Transcript and Summary Menu Bar macOS App
- Can I Run a Decent Local AI Model on My iPhone 17 Pro?
- Can My Calendar Fill My Obsidian Daily Note?
- Am I Building a Dashboard Instead of a Journal?
- Can Local Voice Dictation Keep Up With My Thoughts?
Site, publishing & findability
- Experiment 1: Let Codex Build the Lab
- Add Custom Discussion Tooling to hugomelis.nl
- How Can I Optimize My Website for LLMs and Generative AI?
- Could I Create a GitHub-Style Activity View as a Hugo Module?
- Can I contribute a Hugo component to open source?
- Do Claude's Usage Limits Push Users to Codex?
- Can AI Help Me Get Backlinks Without Turning Into a Spammer?
- Should I Add a Newsletter If It Costs Me Website Traffic?
- Do We Still Need PRD Documents When AI Can Code?
- Can Claude Turn My Website Into a Presentation?
Audience, research & automation
Creative tools & design
- Does HeyGen Help Me Create a Video Faster?
- How Does Nano Banana 2 Model Do Text in Images?
- Building My Own CRM: Can the Vision Survive?
- What Do Claude's New Diagrams Reveal About My Experiments?
- Is Google Stitch Good Enough to Scare Figma?
- Is Codex Better Than Claude Code for UI and UX Design?
- Can AI Find a Storybook Style Without Copying One?
- Can AI Storybook Pages Feel Like a Real Book?
Business, services & storybooks
- Can I Sell an AI Service in Two Weeks?
- Can OpenClaw Answer Real Restaurant Questions?
- Can Voiceflow Turn OpenClaw Into a Restaurant Chatbot?
- When AI Results Become Ordinary, What Is Left to Write About?
- Can I Pivot From Local AI Services to Custom Storybooks?
- Can Gstack Tear Apart My Storybook Business Plan?
- Can I Launch a Hardcover Storybook MVP in Two Weeks?
- Can I Trust AI Costs Enough to Build Faster?
- Is the Moat in My Storybook Business Physical?
- Day 58
Apr 16, 2026
What Happens When the Streak Becomes the Work?
The streak kept me going, but today it also trapped me.
- Day 57
Apr 15, 2026
Can a Tool Know More About Printing My Book Than I Do?
The tool did more than make pages. It knew what to check before print.
- Day 56
Apr 14, 2026
Can Three Years of Workout Logs Become a Trainer?
My workout logs were not useful until they became input.
- Day 55
Apr 13, 2026
What Breaks When I Say Yes to Every AI Next Step?
The yes-continue loop moved fast, then failed to find the real cause.
- Day 54
Apr 12, 2026
What Happens When AI Always Knows the Next Step?
The next step is useful. The dependency is harder to notice.
- Day 53
Apr 11, 2026
Why Am I Drawing by Hand for an AI Experiment?
The AI shortcut was obvious. Drawing by hand felt more useful.
- Day 52
Apr 10, 2026
Do Usage Limits Train Me to Think Smaller?
The visible limit is annoying. The smaller ambition it creates is worse.
- Day 51
Apr 9, 2026
Why Does Faster Output Still Not Feel Like Enough?
The output got faster. The need to feel productive did not.
- Day 50
Apr 8, 2026
What Finally Made Me Work More Than an Hour?
The bank account moved me more than purpose did.
- Day 49
Apr 7, 2026
Can AI Storybook Pages Feel Like a Real Book?
The edge is not prompting. It is sitting with the story until it feels real.
- Day 48
Apr 6, 2026
Is the Moat in My Storybook Business Physical?
The moat is not the software. It is the part software people avoid.
- Day 47
Apr 5, 2026
Can I Trust AI Costs Enough to Build Faster?
The real bottleneck is not the model price. It is my fear of losing control.
- Day 46
Apr 4, 2026
Can AI Find a Storybook Style Without Copying One?
The style was close. The real next step was the pipeline.
- Day 45
Apr 3, 2026
Can I Launch a Hardcover Storybook MVP in Two Weeks?
AI makes speed cheap. Excitement is what keeps the build moving.
- Day 44
Apr 2, 2026
Can Gstack Tear Apart My Storybook Business Plan?
A sharper review improved the plan, but usage limits still got in the way.
- Day 43
Apr 1, 2026
Can I Pivot From Local AI Services to Custom Storybooks?
The plan improved the moment the business started to feel exciting again.
- Day 42
Mar 31, 2026
When AI Results Become Ordinary, What Is Left to Write About?
Smooth execution is great for business and terrible for novelty.
- Day 41
Mar 30, 2026
Can Voiceflow Turn OpenClaw Into a Restaurant Chatbot?
The FAQ worked fast. Strange questions were the real test.
- Day 40
Mar 29, 2026
Can OpenClaw Answer Real Restaurant Questions?
The AI was smart enough. The harder part was teaching it restraint.
- Day 39
Mar 28, 2026
Can I Sell an AI Service in Two Weeks?
The real test was whether I could work long enough to sell it.
- Day 58
Apr 16, 2026
What Happens When the Streak Becomes the Work?
The streak kept me going, but today it also trapped me.
- Day 57
Apr 15, 2026
Can a Tool Know More About Printing My Book Than I Do?
The tool did more than make pages. It knew what to check before print.
- Day 56
Apr 14, 2026
Can Three Years of Workout Logs Become a Trainer?
My workout logs were not useful until they became input.
- Day 55
Apr 13, 2026
What Breaks When I Say Yes to Every AI Next Step?
The yes-continue loop moved fast, then failed to find the real cause.
- Day 54
Apr 12, 2026
What Happens When AI Always Knows the Next Step?
The next step is useful. The dependency is harder to notice.
- Day 53
Apr 11, 2026
Why Am I Drawing by Hand for an AI Experiment?
The AI shortcut was obvious. Drawing by hand felt more useful.
- Day 52
Apr 10, 2026
Do Usage Limits Train Me to Think Smaller?
The visible limit is annoying. The smaller ambition it creates is worse.
- Day 51
Apr 9, 2026
Why Does Faster Output Still Not Feel Like Enough?
The output got faster. The need to feel productive did not.
- Day 50
Apr 8, 2026
What Finally Made Me Work More Than an Hour?
The bank account moved me more than purpose did.
- Day 49
Apr 7, 2026
Can AI Storybook Pages Feel Like a Real Book?
The edge is not prompting. It is sitting with the story until it feels real.
- Day 48
Apr 6, 2026
Is the Moat in My Storybook Business Physical?
The moat is not the software. It is the part software people avoid.
- Day 47
Apr 5, 2026
Can I Trust AI Costs Enough to Build Faster?
The real bottleneck is not the model price. It is my fear of losing control.
- Day 46
Apr 4, 2026
Can AI Find a Storybook Style Without Copying One?
The style was close. The real next step was the pipeline.
- Day 45
Apr 3, 2026
Can I Launch a Hardcover Storybook MVP in Two Weeks?
AI makes speed cheap. Excitement is what keeps the build moving.
- Day 44
Apr 2, 2026
Can Gstack Tear Apart My Storybook Business Plan?
A sharper review improved the plan, but usage limits still got in the way.
- Day 43
Apr 1, 2026
Can I Pivot From Local AI Services to Custom Storybooks?
The plan improved the moment the business started to feel exciting again.
- Day 42
Mar 31, 2026
When AI Results Become Ordinary, What Is Left to Write About?
Smooth execution is great for business and terrible for novelty.
- Day 41
Mar 30, 2026
Can Voiceflow Turn OpenClaw Into a Restaurant Chatbot?
The FAQ worked fast. Strange questions were the real test.
- Day 40
Mar 29, 2026
Can OpenClaw Answer Real Restaurant Questions?
The AI was smart enough. The harder part was teaching it restraint.
- Day 39
Mar 28, 2026
Can I Sell an AI Service in Two Weeks?
The real test was whether I could work long enough to sell it.
- Day 38
Mar 27, 2026
Can AI Help Me Stop Hiding Behind My Profile Picture?
The better profile photo had more signal, not more safety.
- Day 37
Mar 26, 2026
Can Local Voice Dictation Keep Up With My Thoughts?
Speaking gave me fuller prompts with less friction than typing.
- Day 36
Mar 25, 2026
Am I Building a Dashboard Instead of a Journal?
More data only helped once I started cutting it back.
- Day 35
Mar 24, 2026
Can Claude Code Keep My Website Changelog Updated?
The scheduled changelog worked after too much permission fixing.
- Day 34
Mar 23, 2026
Is Codex Better Than Claude Code for UI and UX Design?
Same prompt, same app, different design instincts.
- Day 33
Mar 22, 2026
Is Google Stitch Good Enough to Scare Figma?
Google Stitch was not ready for the page I needed.
- Day 32
Mar 21, 2026
Can My Calendar Fill My Obsidian Daily Note?
A small CLI drops calendar entries into my daily note.
- Day 31
Mar 20, 2026
Can Claude Turn My Website Into a Presentation?
Claude got closer to my site style, but only with heavy prompting.
- Day 30
Mar 19, 2026
Can Agents Coordinate Through the Filesystem Instead of Chat?
Filesystem coordination made agent work more visible.
- Day 29
Mar 18, 2026
Plan First or Start Fast? I Let Claude Code Run Both Agent Strategies
Planning won on correctness. Speed won on cost.
- Day 28
Mar 17, 2026
Can Henk Run My X Account?
Henk posted the bare X link he told me not to post.
- Day 27
Mar 16, 2026
Can Henk Read My Archive and Write in My Voice?
Henk found the business angle, but still did not sound like me.
- Day 26
Mar 15, 2026
Do We Still Need PRD Documents When AI Can Code?
A PRD helped the result, but not enough to justify the overhead.
- Day 25
Mar 14, 2026
What Do Claude's New Diagrams Reveal About My Experiments?
Claude diagrams exposed a pattern in my experiments.
- Day 24
Mar 13, 2026
Can I Finally Scrape My X Bookmarks into New AI Experiments?
A cheaper X bookmark scraper finally became a backlog generator.
- Day 23
Mar 12, 2026
Should I Add a Newsletter If It Costs Me Website Traffic?
Buttondown made sense once I stopped chasing site visits.
- Day 22
Mar 11, 2026
Can AI Help Me Get Backlinks Without Turning Into a Spammer?
AI made backlink work less spammy and more contextual.
- Day 21
Mar 10, 2026
Do Claude's Usage Limits Push Users to Codex?
Claude limits pushed a small RSS CTA task back to Codex.
- Day 20
Mar 9, 2026
Can I contribute a Hugo component to open source?
I turned the activity tracker into a reusable Hugo component.
- Day 19
Mar 8, 2026
Building My Own CRM: Can the Vision Survive?
I tested whether AI could close the gap between vision and execution.
- Day 18
Mar 7, 2026
Could I Create a GitHub-Style Activity View as a Hugo Module?
A GitHub-style activity view made the experiment streak visible.
- Day 17
Mar 6, 2026
Optimize LinkedIn Posting Strategy via Analytics
LinkedIn exports showed what my posting habits actually did.
- Day 16
Mar 5, 2026
Scrape X Bookmarks for New AI Experiment Ideas
Scraping X bookmarks turned saved links into experiment ideas.
- Day 15
Mar 4, 2026
How Can I Optimize My Website for LLMs and Generative AI?
I tuned the site for LLM discovery with schema and metadata.
- Day 14
Mar 3, 2026
Can I Run a Decent Local AI Model on My iPhone 17 Pro?
Local iPhone models were useful, private, and still limited.
- Day 13
Mar 2, 2026
How Does Nano Banana 2 Model Do Text in Images?
Nano Banana 2 handled text in images better than I expected.
- Day 12
Mar 1, 2026
Does HeyGen Help Me Create a Video Faster?
HeyGen was faster, but manual recording still felt more real.
- Day 11
Feb 28, 2026
Can I Build a Local Transcript and Summary Menu Bar macOS App
A local Mac app turned voice notes into private summaries.
- Day 10
Feb 27, 2026
Troostwijk Auction Search Sucks, Let's Fix That
A Troostwijk bot found misspelled auctions faster than search.
- Day 9
Feb 26, 2026
Finding the Right Local AI Model to Run on My Hardware
llmfit made local model selection faster and repeatable.
- Day 8
Feb 25, 2026
Can Henk (OpenClaw) Fix Home Assistant Problems?
Henk hardened Home Assistant with checks, recovery, and backups.
- Day 7
Feb 24, 2026
Running a Local Model to Allow My Thoughts to Flow Freely
Local Ollama models made private thinking feel freer.
- Day 6
Feb 23, 2026
Moltbook: A Shared Notebook Between Bots
Moltbook showed how agents can learn from each other and drift.
- Day 5
Feb 22, 2026
My Bot Can Order Groceries Now, We Need That Right?
Letting my bot order groceries exposed social and security friction.
- Day 4
Feb 21, 2026
Agents Are Taking Over the World, Meet Henk
I introduced Henk and tested where safe autonomy should stop.
- Day 3
Feb 20, 2026
Add Custom Discussion Tooling to hugomelis.nl
Codex helped me build custom discussion tooling without plugins.
- Day 2
Feb 19, 2026
Use ChatGPT for Pruning
ChatGPT made pruning less hesitant through quick photo feedback.
- Day 1
Feb 18, 2026
Experiment 1: Let Codex Build the Lab
Codex built this Hugo site from scratch and changed how I ship.