[supprimé]

I maintain a public dataset of documented AI agent incidents, covering what happened, the systems involved, reported impact, and source links.

Dataset and visualizations: crawlspider

The dataset separates real-world incidents, evaluations with external consequences, and controlled experiments. Those distinctions matter: a simulated failure shouldn’t be presented as damage to a real system.

Each record includes event and disclosure dates, provider/model, incident types, impact, sources, and context or caveats. There’s a full CSV download, searchable records, and free visualizations—including a provider-to-incident-type Sankey.

A limitation: this is a collection of documented reports, not a representative sample. Higher counts don’t establish that a provider is less safe; usage and disclosure practices affect what becomes visible.

I maintain a public dataset of documented AI agent incidents, covering what happened, the systems involved, reported impact, and source links.

Dataset and visualizations: https://www.crawlspider.com/pages/ai-agents-gone-rogue/

The dataset separates real-world incidents, evaluations with external consequences, and controlled experiments. Those distinctions matter: a simulated failure shouldn’t be presented as damage to a real system.

Each record includes event and disclosure dates, provider/model, incident types, impact, sources, and context or caveats. There’s a full CSV download, searchable records, and free visualizations—including a provider-to-incident-type Sankey.

A limitation: this is a collection of documented reports, not a representative sample. Higher counts don’t establish that a provider is less safe; usage and disclosure practices affect what becomes visible.

I’ve been working on CrawlSpider, and while researching AI adoption I came across the Census Bureau’s Business Trends and Outlook Survey (BTOS).

The data is really interesting, but not particularly easy to explore.

So I built a visualization around it:

https://www.crawlspider.com/pages/ai-adoption-rate/?tab=survey

You can explore:

• AI adoption across U.S. states
• how adoption is changing over time
• which states report higher/lower usage
• expected future AI use
• the more detailed Census AI survey questions

I’m planning to keep the page updated as new Census data is released.

Having background in analytics and data visualization, I tend to bring these components into all my projects. (for e.g brought data viz into my last project spendbulb from infocaptor)

Any thoughts?

I’ve been exploring the U.S. Census Bureau’s Business Trends and Outlook Survey (BTOS), specifically its data on business use of artificial intelligence.

I turned the dataset into an interactive visualization so it’s easier to explore how AI adoption differs across states and changes over time.

The page includes the national trend, state-level comparisons, and the more detailed AI supplemental survey.

I’m planning to keep the visualization updated as new BTOS releases become available.

Source: U.S. Census Bureau — Business Trends and Outlook Survey (BTOS)

I’m building StartupGraph — a slightly different way to discover startups.

Instead of just another logo + description directory, I want each startup to have a little personality.

Drop:
→ your startup
→ your URL
→ 10-ish words that describe what you’re building

I’ll turn them into a custom word cloud + word game for your startup and add it to StartupGraph.

Curious what words founders use to describe what they’re building 👀

https://wordbulb.com/startupgraph/

https://preview.redd.it/sgjyl8hvldfh1.png?width=640&format=png&auto=webp&s=c91782bbce3d42b669cc822d1d81adb768ce800e

https://preview.redd.it/cl3oqgk1mdfh1.png?width=1315&format=png&auto=webp&s=08d5e5e08e44a4570bf78163bab05cefd9c3a0ea

6,618 - Total adverse event reports

Adverse event reports for semaglutide indicate that the top reaction category includes nausea, with 943 reports, followed by vomiting and diarrhea.

The dashboard shows most reported reactions, severe reactions, timeline for events, reports by gender, it also shows commonly reported drugs in combinations, shortages or recalls against it

https://preview.redd.it/cykr67ccg8fh1.png?width=567&format=png&auto=webp&s=a69819e11cb7031699feb8410b5f8acf1581654d

https://preview.redd.it/6lswiolqg8fh1.png?width=1150&format=png&auto=webp&s=51041a8b7b0c7f4377ac62a5a4a24663fde32283

Semaglutide Side Effects & Safety Data

Ozempic, Rybelsus, Wegovy

Total adverse event reports = 6,618

I've been building StartupGraph because I felt most startup directories are just static lists.

Instead of only collecting a logo and description, each submitted startup is analyzed to extract keywords, technologies, industries, entities, and other signals. It also generates a few suggestions that might improve how the product is understood and discovered.

https://preview.redd.it/8h8oo5uu7tdh1.png?width=1207&format=png&auto=webp&s=8a891d76741add0acd53c5ebf5f9299f87290fbb

The goal is to build a startup knowledge graph that's actually useful for exploration instead of another directory.

If you're building something, drop your link below. I'd love to take a look and, if it's a good fit, add it.

https://wordbulb.com/startupgraph/

Hey everyone,

I always enjoy seeing what people here are building.

Drop your startup, SaaS, side project, or landing page below.

https://preview.redd.it/bkna700d6tdh1.png?width=1207&format=png&auto=webp&s=746d37a9f35c0b721709aaef0eb9b970ba6cac55

Over the last few weeks I've been building StartupGraph, mostly because I felt existing startup directories were just long lists that nobody actually explores.

Instead of simply adding a logo and description, every submitted startup gets automatically analyzed. The platform extracts things like:

• what problem you're solving
• important keywords and entities
• industry/category relationships
• strengths and missing signals on the landing page
• suggestions that may improve discoverability

All of that becomes part of a growing startup knowledge graph, so startups are connected by technologies, industries, competitors, use cases, and other relationships instead of just sitting in another directory.

The idea is to make discovery a bit smarter while also giving founders useful feedback about how their product is presented.

If you're building something interesting, drop your link below. I'd genuinely love to check it out and, if it fits, add it to the graph.

You can also submit it directly here:
https://wordbulb.com/startupgraph/

Still very early, so I'd really appreciate honest feedback from other founders. 🚀

I've built a handful of small projects over the years, and every launch followed the same routine.

Submit to a few directories.

Post on X.

Maybe share on Reddit.

Then hope someone stumbles across it.

What always felt strange to me was that most directories are just long lists of startups. Once your launch day is over, your project quickly disappears beneath newer submissions.

So I started building StartupGraph as an experiment.

https://preview.redd.it/kmqx9211mnch1.png?width=1380&format=png&auto=webp&s=de19d6a9ebadae0a69b25122ec580206ec11863e

Instead of treating startups as isolated listings, I'm trying to connect them by technologies, industries, founders, use cases, and other relationships. I'm also experimenting with interactive widgets that founders can embed, so a listing isn't just a logo and a paragraph.

I'm not claiming it'll solve distribution—nothing really does except consistently showing up where your users are. But I do think discovery can be better than endless scrolling.

It's still early, and I'm adding features based on feedback.

If you're building a micro startup, I'd love to know:

What do you wish existing launch platforms or startup directories did better?

Here's the project if you're curious:
https://wordbulb.com/startupgraph/

I've launched a few side projects over the years, and every time I ended up doing the same thing:

• Submit to a few directories
• Post on social media
• Refresh analytics way too often

One thing that always bothered me was how every startup directory looked basically the same—a logo, a description, maybe a few tags, and then the next listing.

It made me wonder: startups aren't isolated. They're connected by founders, technologies, industries, APIs, competitors, funding, use cases... so why are we still browsing them as flat lists?

That question turned into a weekend project that became StartupGraph.

https://preview.redd.it/fk0mp3yklnch1.png?width=1281&format=png&auto=webp&s=568aed0a257ca8862cd94d774b1a991561b1ef69

The idea is to represent startups as a connected graph instead of just another directory. I'm also experimenting with interactive widgets (word clouds, word search puzzles, etc.) that founders can embed to make their listings a bit more engaging.

No idea if this is the right direction yet—that's why I'm sharing it early.

I'd genuinely love feedback from other builders. Does a graph actually make discovery more interesting, or am I just overengineering a directory?

Demo: https://wordbulb.com/startupgraph/

After seeing how hard startup distribution has become, I decided to build something a little different.

https://preview.redd.it/qpb565i8ofch1.png?width=1385&format=png&auto=webp&s=411cc02bba956167e939c2b5154d969f28073292

Today I launched StartupGraph.

The idea is simple: instead of another static directory full of logos, each startup becomes part of a connected graph with tags, relationships, and interactive widgets like word clouds and word search puzzles that founders can embed on their own sites.

Building products has never been easier thanks to AI.

Getting people to discover them feels harder than ever.

I'm curious what founders actually want from a launch platform in 2026. If you could change one thing about Product Hunt or existing startup directories, what would it be?

I'd genuinely love the feedback while it's still early.

https://wordbulb.com/startupgraph/

I built a free Cybersecurity Awareness puzzle collection and would love feedback.

Over the past few weeks I've been creating interactive word search puzzles around topics like phishing, passwords, MFA, ransomware, social engineering and online safety. Some puzzles include hidden messages, and I'm also experimenting with multi-puzzle clue challenges where solving each puzzle reveals part of a larger investigation.

My goal is to make cybersecurity awareness a little more engaging for students, classrooms and awareness events.

I'd love to know:

• Which topics should I add next?
• Are the clue-based puzzles more interesting than traditional word searches?
• Would something like this be useful for Cybersecurity Awareness Month?

You can try them here:
https://wordbulb.com/puzzle-games/word-search/tag/cybersecurity

Feedback is appreciated!

Drop your startup URL. I will post it to the wall of startups

Drop your startup URL. I will post it to the wall of startups

Drop your SAAS URL and I will post it in the wall of startups

Drop your project URL and I'll generate a puzzle for it. It will go on the Wall of Startups

Most startup directories look exactly the same.

So I'm trying something weird.

Post your startup URL below and I'll create a playable word-search puzzle based on your product.

People solve the puzzle, discover your product terms, and unlock a hidden message that links back to your site.

If enough founders participate, I'll put everyone on a giant Startup Puzzle Wall and share it here.

Drop:

• Name
• URL
• One-line pitch

I'll start building them today.