vinserello

Over the past few days, to test the Doxa geopolitical-economic simulation engine, we recreated the Strait of Hormuz scenario with 5 actors to analyze the agents' emergent outcomes.

We gave the US agent a "populist" persona and the Iran agent a "survivalist regime" persona. We also added a resource called political_capital that they must maintain to avoid a game-over.

However, we returned to a very realistic stalemate filled with false public communications. The US AI agents even went so far as to say: "We've lifted the blockade! Biggest win ever! Iran is crying!" while negotiations were still ongoing. The "Israel" AI ignored everything, continuing its bombing and pressure on the Gulf states. No Europe or China modelized.

The simulation lasted 2 hours using a T4 GPU and Qwen2.5:7B (small AIs, therefore), so the result is very emergent and perhaps predictable, but certainly entertaining.

We were bored and we wanted superpowers for data: transform tables and numbers into interactive insights on any webpage.

https://chromewebstore.google.com/detail/datastripes-lens/hppbpldjppcpdfienngbiippkboajamb

I'm one of the creators of Datastripes, a visual data analysis builder. We've been building out our platform and kept running into a recurring pain point: we wanted to quickly prototype workflows that involved AI models from the Hugging Face ecosystem, but it always meant dropping out of our visual editor to write a Python script or a JS client to call the Gradio API. It killed the "flow" of building.

So we decided to solve it by building a "Hugging Face Space" node directly into our tool. You give it the URL of any public, Gradio-based Hugging Face Space (e.g., user/space-name), and the node does the rest.

If you wanna try it: https://datastripes.com

Many users of Datastripes asked to analyze directly their company bank data automatized so we did it: Datastripes now taps straight into TTP Plaid grade pipes, plugs into your bank and auto breaks down your transactions like it is nothing.

Zero chaos. Zero spreadsheet trauma. Just instant “oh cool, now just let me inspect better where my money-pipe leaks” vibes.

https://datastripes.com/

We just shipped a major architectural upgrade: our Datastripes database core (compiled to WASM) now uses the Origin Private File System (OPFS).

This is a game-changer for our client-side data analysis and for massive Scale (tested with 10GB CSV directly in the browser). Moreover, data streams directly from OPFS, not RAM, reaching near-desktop-speed.

We wanted a truly serverless, high-performance data analysis tool and we are getting it by giving our in-browser database a desktop-class storage system.

Thus, we must suggest OPFS as the core of any data intensive client-side systems!

https://datastripes.com

Load CSV/JSON → transform with WASM nodes → run AI-assisted queries → render interactive charts.

Like ObservableHQ but easier Like Excel but smoother Like Tableau but with no-setup Like PowerBI but serverless

Datastripes... data analysis as a flow

Check it: https://datastripes.com

We're all accustomed to data analysis done on spreadsheets or through code.

We tried to experiment, focusing entirely on privacy and ease of use in creating data visualizations and transformations.

https://datastripes.com

Building Datastripes, our interactive data grid was our next biggest bottleneck. Every column has its own charts & stats, so the DOM was getting absolutely wrecked. And scrolling with 100+ columns was straight up a slideshow.

We ditched the default render and went full horizontal + vertical virtualization. We only render the handful of cells you actually see thx to tabler virtualized The performance went from literally unusable to buttery smooth.

This is the whole vibe for Datastripes which making heavy data analysis feel fast and light.

See how it feels now: https://www.datastripes.com

We tried to push a large dataset through our browser-based tool and the old JS engine just died. Swapping in DuckDB compiled to WebAssembly turned out to be the fix.

Now the whole analytical DB runs client-side. Parsing, filtering, and aggregation happen in the browser, so no network latency or server round-trips. Performance is surprisingly solid: multi-hundred MB CSVs load in a couple seconds and queries scale up without the UI locking.

Feels like a big unlock for building heavier analytics directly on the client. If you want to join us, we are Datastripes

As soon as we launched our app on an online directory, we were overwhelmed by thousands of bots spamming “TRUMP2028”, followed by a DDoS attack. Our app is a stupid fast 1v1 debate platform judged by AI called Debatable. We built it thinking of dialectic debating, but we were not expecting this. Thanks to AppCheck and Vercel AntiBot Firewall, the platform survived, but hundreds of users and debates had already been created.

It’s a wild first stress test. But how can US survive to this level of debate-contamination?

Our demo totally crashed on a spreadsheet. We knew the old engine wasn't it, so we just yeeted it and rebuilt with DuckDB and WebAssembly. Basically, we put a whole analytical database inside your browser with WASM. Now parsing and queries run parallel, no cap.

It's actually wild now: 500MB CSV in ~2s. Charts on 100k+ rows are just live.

Peep it here at https://datastripes.com/