How far would you go to save $36.99?
Turns out that for me, the answer is that I'd go quite far.
When I was younger, I'd memorised all the tricks for getting free wi-fi out of paid wi-fi. Append ?image.jpg to a captive portal to get around it, custom DNS servers, a VPN listening on TCP 53, that sort of thing.
Sure, I could have paid $1 and been online, but the paywall was an actual affront that required I work for however long it took to build however brittle a solution to bypass it.
So when I see a system like United's Wi-Fi, that only lets you use iMessage and WhatsApp for free like some kind of luddite, it gets me thinking.
A bit of backstory
I like to think I'm at the forefront of technology, and have been pretty much since I picked up my first laptop (read: my mum bought it for me) and started installing packages, writing scripts, and staying on top of the latest news.
Over the past year or two it's all moved faster than ever, and keeping up takes a genuinely concerted effort. Even a couple of weeks away on vacation means extra reading on return.
Earlier this year I used OpenClaw to build Lawrence Craw, my personal assistant. Larry had his own email on my domain, his own phone and number, and an anti-bot-detection browser at his disposal.
With a lot of coaxing and a painfully curated TOOLS.md, I got him into a few WhatsApp groups and even had him book me dinner.
Walking into the restaurant and giving the name Lawrence was a genuine TRIP.
But I never really found joy in it. I was babysitting constantly. The memory didn't really work, I couldn't share Larry with my friends, and every slightly advanced request meant a pile of re-prompting and a pile of tokens.
One thing stuck though: I could talk to Larry over WhatsApp mid-flight. I'd ask him to track my flight, work out whether I'd make my connection, suggest restaurants at my destination. But with no real memory or any ability to plan before executing, I mostly got the AI equivalent of a magic 8 ball:
"my sources say maybe."
It was never the intelligence
I'd worked this out a while ago: it's not an intelligence or execution problem anymore. The models are smart enough to get things done, and execution is easy when you've got a command line and a browser. The real challenge is the same as it's always been: context.
So for this trip I tried something different. I'd started using PromptQL, and the thing that actually changed the game was that it could hold context across the whole plan, even with friends in on it.
I planned the entire trip with it as a partner from inception: bouncing ideas, checking my email, curating the calendar, grinding through the back-and-forth I normally can't be bothered with.
If it was in on the plan from the start, it stood a real chance of being useful when things shifted mid-trip.
And it was. I wasn't babysitting, I had an actual travel agent who knew everything about me.
Anyway, back to the paywall
Thinking forward to my long flight to Europe and the Wi-Fi about to burn a $36.99 shaped hole in my pocket, I wondered how I'd be able to keep planning, exchanging emails, and chasing recommendations the whole way over.
Then I thought... WWLD (What Would Lawrence Craw Do).
United already gives me WhatsApp for free.
If WhatsApp could talk to PromptQL, I'd get exactly what I wanted and for the joyous price of $0
Thus, PontQL was born. Pont is French for bridge (Welsh too apparently). I was going to France and bridging WhatsApp and PromptQL, so it all felt very appropriate.
With PontQL running on my server (behind iptables you animals) I put in my phone number & PromptQL PAT and scanned the QR code.
Team AI with a wiki... via WhatsApp.
If you want to play with it
PontQL is deliberately simple:
- Download
- Run (docker or npm)
- Enter your number & PAT
- Scan the QR code
- ?????
- Profit
PromptQL hands back tables, charts and dashboards, rendered as image attachments in WhatsApp and offers to learn continuously to hydrate the wiki and build up its own brain.
It feels like OpenClaw that actually works which is honestly what I'm here for