Can I blow your mind with some economic indicators?

I spend a lot of time looking at economic data. I’m constantly pulling numbers from Statistics Canada, the OECD, FRED, the Bank of Canada trying to get a clear picture of how the Canadian economy is actually performing.

The problem is that getting those numbers is usually very annoying.

Do you want to know Canada’s quarterly GDP growth? Go to Statistics Canada. Do you want to compare it to the United States? Call FRED.

Do you want to see where both countries rank among the OECD? That’s a different website with a different interface.

Do you want to check the household debt-to-income ratio, the output gap, and the fiscal balance in the same sitting? Soon you’ll find yourself opening a dozen tabs, cross-referencing methodologies, and hoping the date ranges line up.

All of the data is public. As a taxpayer, you funded its collection. But accessing it in any coherent way has usually required either a Bloomberg terminal or a masochistic desire.

I wanted something better. So I built it.

And I got a neat domain name for it too! It’s at economy.tools

economy.tools is a free, open dashboard that tracks 109 economic indicators for Canada and the United States. Every number comes directly from an official source. No opinions – the data speaks for itself.

This all comes from seven sources:

  • OECD – the backbone, providing internationally comparable data across most indicators
  • Statistics Canada – Canadian-specific series like employment surveys and building permits
  • Bank of Canada – interest rates, exchange rates, and monetary policy indicators
  • FRED (Federal Reserve Economic Data) – the best source for US economic series
  • World Bank – structural and demographic indicators
  • Bank for International Settlements – credit, debt, and financial stability data
  • IMF – World Economic Outlook projections

The indicators span ten categories: GDP and growth, inflation, labour markets, fiscal policy, monetary policy and interest rates, housing, trade, productivity,
financial stability, and demographics.

Each one includes a detailed explainer: what it measures, how it’s calculated, and why it matters.

The feature I find most useful is the Canada-US comparison.

You can see at a glance where Canada leads, where it lags, and by how much. Every indicator gets a ranking against OECD peers, a z-score showing how far each country sits from the average, and a side-by-side chart going back years.

This matters because so much of Canadian economic debate happens without reference points.

The dashboard also includes OECD Economic Outlook projections, so you can see not just where things stand but where they’re headed.

I built economy.tools with AI using Claude Code, Anthropic’s AI coding tool. The entire project – data ingestion pipeline, API, frontend, 109 indicator explainers, pre-rendered static site – took about eight hours.

That’s not an exagerration. It took one working day. The architecture is a monorepo with three pieces: a shared package that defines every indicator and its data sources, a backend API that fetches and stores observations from all seven providers, and a React frontend that presents everything in a clean dashboard.

The site loads fast, works without a backend, and is fully indexable by search engines. Could I have built this without AI assistance? I might have given up due to sheer boredom. The tedious parts such wiring up seven different API adapters with different authentication schemes and data formats, writing 109 explainers with LaTeX formulas, building a responsive dashboard with charts and sparklines got done without breaking a sweat with AI. I focused on what the product should do and how the data should be presented.

The implementation moved fast.

I believe that informed citizens make better decisions. That starts with having access to the same data that policymakers and economists use.

Economic data shouldn’t be locked behind expensive terminals or scattered across dozens of government websites. It’s public information. It should be easy to find, easy to read, and easy to compare. My site is free.

If you work in public policy, journalism, research, or you just want to understand what’s happening in the Canadian economy, please enjoy! I’m just happy websites are starting to become useful again.

Are Canadian political polls still useful?

I’ve long viewed polling as a useful but imperfect tool for gauging Canadian sentiment. It helps parties like the Conservatives under Pierre Poilievre create messaging that resonate with everyday Canadians frustrated by Liberal overreach on everything from taxes to immigration.

But lately, I’ve been wondering: has polling on key issues reached a ceiling in its usefulness? In an era of deepening divides and polarization, are these surveys truly illuminating public opinion, or are they just reinforcing the tribal lines we already know exist?

Like in many places, it’s no surprise that Canadian society has also become polarized, but I suspect the root cause isn’t purely ideological. It’s may just be our partisan nature.

Our evolutionary brains, wired for survival in tribes, prioritize loyalty to “our side” over nuanced contemplation of ideas. We gravitate toward who we like or oppose, rather than dissecting why we support or reject certain policies.

This tribalism explains why debates on climate change, immigration, and media trust often feel like proxy wars for team allegiance rather than genuine discourse. Conservatives rally around common-sense critiques of elite-driven narratives, while Liberals defend expansive government interventions. But in this setup, polling might be capturing signals of partisan identity more than authentic beliefs.

Inside the Abacus Data Poll – What It Attempts to Reveal About the Conservative Base

Take a recent poll by Abacus Data, conducted January 12-16, 2026, surveying 2,008 Canadian adults as part of a series with the Toronto Star ahead of the Conservative Party convention. It explores the Conservative base and how it stacks up against the rest of the country on hot-button issues. The findings highlight the limitations of traditional polling, in my opinion. Statistical tools are useful if they ask the right questions.

On climate change, for instance, only 34% of the Conservative base sees it as an urgent problem demanding immediate action, compared to 69% of everyone else and 58% overall. Instead, 73% of the base prioritizes affordability and energy security over accelerating green initiatives, versus 53% of non-base respondents. And 72% support approving more major energy projects like pipelines, against 56% of others.

From a conservative perspective, this makes sense: we’re pragmatic about balancing environmental concerns with economic realities, especially when “net-zero” mandates from Ottawa threaten jobs in Alberta’s oil patch or hike heating bills in Ontario.

But here’s the problem: these questions are heavily partisan-coded. Anyone with even a passing knowledge of Canadian politics knows climate alarmism is a left-wing staple, pushed by Liberals and NDP.

The Hidden Bias: How Partisan-Coded Questions Distort Real Opinions

Respondents might be answering not based on personal conviction, but to align with their team and perceived value. It’s reminiscent of those high school career surveys where you’d tick boxes you know were coded for “doctor” because that job is prestigious, not because you actually enjoyed biology. Whereas most of us would answer in the affirmative to “Do you like being outside?” if we were being honest, but didn’t because we didn’t want to dig ditches.

“Do you believe climate change is an existential threat?” Sure, if you’re signaling Liberal virtue. But are they your actual values, or the confort of your tribe?

Immigration and integration questions reveal similar coding. The poll shows 74% of the Conservative base favoring severe reductions in immigration levels, compared to 57% of others, with 63% of all Canadians agreeing on major cuts. Even starker, 78% of the base believes Canada has gone too far in accommodating different values and cultures, versus 49% of non-base Canadians.

Conservatives have been vocal about this. Poilievre’s push for controlled borders resonates because unchecked immigration strains housing, healthcare, and social cohesion. Yet these queries scream “right-coded.” They tap into known partisan divides: Liberals champion multiculturalism as an unqualified good, while conservatives worry about integration failures.

Poll respondents, sensing the “answer key,” might default to their tribe’s position rather than reflecting on personal experiences, like whether their community has benefited from or been burdened by rapid influxes.

Media trust is another flashpoint, and perhaps the most right-coded of all.

Only 23% of the Conservative base believes mainstream media reports facts fairly, compared to 62% of everyone else. Meanwhile, 69% of the base sees the system as rigged for elites, versus 48% of others, and just 26% trust institutions like courts or Elections Canada to function fairly, against 60% of non-base respondents.

As a conservative, this skepticism rings true. The CBC’s Liberal-friendly bias and Big Tech’s censorship of dissenting voices have eroded faith in legacy outlets. But again, these questions are loaded: media distrust is understood as right-wing populism, from Trump south of the border to Poilievre’s “gatekeepers” rhetoric. People know the script and recite it accordingly.

Rethinking the Questions – Probing Deeper Motivations

This partisan coding is a symptom of polling’s diminishing returns in an age of disruption – societal and technological. Old dividing lines such as left vs. right on climate or immigrationare being shaken by global events like supply chain crises, AI-driven job shifts, and cultural upheavals.

Traditional surveys feel dated, inefficient at probing deeper motivations without tipping off the “correct” answer. What if, instead of asking “Do you support carbon taxes?” (left-coded yes), we posed questions like “How do you weigh personal financial stability against long-term environmental goals?”

Or on immigration: “What community changes have you noticed in the last five years, and how do they affect your daily life?” These indirect probes could reveal fundamental drivers – fear of economic insecurity, desire for continuity – without cueing tribal responses.

We might discover that many Canadians, regardless of party, share conservative instincts on fiscal prudence or border security, but express them differently.

AI and Embeddings: The Future of Understanding Canadian Public Opinion?

This is where modern innovations like AI and large language models (LLMs) could outshine traditional pollsters.

AI offers tools to map concepts in ways that bypass human biases and scripted answers. At their core, LLMs use “embeddings” which are mathematical representations of words, phrases, or ideas as high-dimensional vectors. These vectors capture semantic relationships: “climate change” might cluster near “environmental policy” and “energy costs,” but AI can quantify similarities without assuming partisan priors.

For example, by analyzing open-ended responses or social media chatter through vector similarity searches, we could match concepts like “affordability” to “immigration pressures” or “media trust” to “elite capture,” revealing hidden connections that polls miss. Concept matching via cosine similarity (a measure of vector alignment) could uncover that a respondent’s skepticism of climate urgency stems not from denialism, but from vector proximity to “government overreach” or “economic hardship.”

We can supplement the experienced insight from pollsters, adding these semantic discoveries to understand motivations beyond tribal signals – and maybe even correct for them.

In the end, as Canada heads toward another election cycle, conservatives should push for smarter insights to cut through the noise. Current polling methods may have hit their useful ceiling. By embracing AI-assisted analysis and fewer coded questions, we might reconfigure a political environment that’s more contemplative and less combative, aligning with the common-sense approaches that puts Canadians first.

Announcing PDFPony.com

I didn’t set out to build a PDF tool.

It started with a simple thought: modern browsers can edit video, run serious WebAssembly apps, and render 3D worlds. So why is “merge two PDFs” still treated like something that needs to happen on a server somewhere?

Around Black Friday I finally cancelled my Adobe subscription. It was pushing about $100/month, and year over year Adobe has been getting more aggressive with policies and product decisions I didn’t love. At a certain point, it stopped feeling like software and started feeling like a toll.

And the alternatives weren’t much better.

If I needed to do a quick task – merge, compress, split, sign – uploading a document to some random website was almost always a non-starter. Not because I had a specific horror story, but because the questions are obvious:

Who owns that server? What country is it in? What’s the retention policy really mean? Who has access? What happens when a “free tool” changes hands?

Plenty of people have had company data exploited or hijacked. You don’t need a personal disaster to recognize a bad default.

So I built a different default.

PDF Pony

PDF Pony is a suite of 140+ PDF tools that runs entirely in your browser.

No uploads. No backend processing. No accounts. No “trust us.” Your files never leave your device because there’s nowhere to send them.

If you want proof, open Chrome DevTools and go to the network tab. You’ll see the site load like any other website, but when you run a tool, you won’t see your PDF being posted to a server. There isn’t one. It’s just a website you download and run.

The promise is simple:

100% privacy and security. I don’t want your documents – you don’t need to send them to me.

This is possible

What I love about the current era of web software is that the ceiling keeps moving. Browsers are real computing environments now. With WebAssembly, Web Workers, and modern JS PDF libraries, local PDF processing is 100% possible.

And honestly, tools like Claude Code (Anthropic doubled my tokens over Christmas – thanks!) have made it easier than ever to explore.

“wait… we can just do this now?”

Turns out, yes.

A feature I thought I’d miss (but didn’t)

The Adobe app I expected to miss most was Adobe Scan.

So I built a PDF camera scanner inside PDF Pony and I think it works better. More importantly, it’s 100% local. No “scan to the cloud,” no account, no silent upload. Just scan then have the PDF on your device.

That one discovery that validated the whole idea for me: the web can deliver “native app” capability without handing your data to someone else.

Some other great tools:

Merge PDFs, Split PDFs, Insert a PDF into a PDF, Flatten a PDF, convert an EPUB into a PDF, compress a PDF, add a QR code to every page of your PDF, add custom stamps to your PDFs, redact text from your PDFs, password protect your PDFs, add a table of contents to your PDF, sign your PDFs, create PDF forms and more!

Speed without servers

Local doesn’t mean slow. I’m sure you’ve got a fast modern computer whose power you underuse 99% of the time.

PDF Pony uses parallel Web Worker processing so heavy operations don’t freeze your browser and it runs faster than you’d expect. Modern machines are powerful. That macbook is powerful. Heck even that Chromebook is. The browser can use that power. The result is a tool that feels instant without relying on infrastructure you can’t see (or don’t own!)

Who it’s for

Anyone who uses PDFs, every day or on occasion.

If you’re a student, a small business owner, an HR person, a lawyer, a political staffer, or just someone trying to get a form done, PDF Pony helps you out. It’s the same set of problems across every profession: “I have a document, I need it slightly different, and I need it done right now.”

A lot of online tools are good products wrapped in a trust problem. PDF Pony is the opposite: the safest workflow is also the easiest workflow. You don’t have to wonder where your file went because it never goes anywhere.

How do I keep the lights on and what’s next?

Right now, hosting is cheap enough that PDF Pony is ad-free, and it’ll stay that way unless traffic becomes overwhelming and I have to start paying some real bills. If I ever add ads, it’ll be because the site is being used so much that it needs the support to keep running.

Next up: more languages, (PDF Pony is now also fully fluent in French, Spanish, German, and Portuguese!) better mobile tooling, and whatever users keep asking for.

Try it and tell me what you think

I want the website to save you money and make less annoyed. That’ll make me happy! If it saves you from paying Adobe for Acrobat or uploading your data to dodgy unknown cloud services, even better.

Use it however you use PDFs, and tell me what you need next.

Email me feature requests at [email protected]