About See AI Work
Welcome to See AI Work (seeaiwork.com)—a no‑fluff notebook of experiments that let anyone watch modern AI earn its hype — or miss the mark.
Why this site exists
Artificial‑intelligence progress is snowballing so quickly that every week feels like a new magic trick. The goal here is to slow things down just enough to answer three plain‑language questions:
- What exactly did the model do?
- How can I check the claim myself?
- What larger pattern does this reveal about where AI is headed?
If a post can’t pass that three‑part test, it doesn’t get published.
What you’ll see here
You get | You will not get |
---|---|
Step‑by‑step breakdowns of live demos (GeoGuessr rounds, code runs, audio tests). | Paywalls, pop‑ups, mailing‑list nagging, or jargon for its own sake. |
Reproducible prompts, datasets, and version numbers. | Blind trust in benchmark leaderboards. |
Sidebars that translate “AI‑speak” into everyday analogies. | Cherry‑picked screenshots with hidden failures cropped out. |
Occasional mistakes, logged and corrected in public. | Affiliate links, sponsored posts, or hidden model fingerprints. |
Who’s writing
I’m Joe—an engineer who measures technology by what it actually ships.
Credentials in 10 seconds:
- AIDS research (the starter kit). I began by building data‑capture and analysis tools for frontline HIV studies—code that passed muster for clinical research, not demo decks.
- Online gaming at scale. Next came helping to build the multiplayer network for a leading game console, keeping millions of players in sync night after night.
- Enterprise data plumbing. I spent the following years shaving latency off a flagship relational‑database engine used across the Fortune 500—because fast queries save real money.
- Blockbuster shooters. Craving player‑facing impact again, I helped wire up the live backbone behind two of the decade’s most‑played sci‑fi shooters, keeping every raid and loot drop ticking.
- AI before it was cool. Long before GPT grabbed headlines, I was running small‑model research projects, figuring out what worked and what was hype.
- AI at industrial scale. Those lessons now power a global customer support operation that fields millions of real customer questions.
Whether it’s netcode, databases, or language models, the method stays the same: pin down the claim, reproduce it, measure the delta, publish the receipts.
How to use this site
- Skim or dive – Each post starts with a one‑paragraph takeaway; everything after that is receipts.
- Run the prompt yourself – Copy‑paste, swap in your own image/text/audio, and compare notes.
- Point out errors – Seen a flaw? Drop a comment or email fix@seeaiwork.com. I’ll update the post with attribution.
Guiding principles
- Clarity beats cleverness – If a sentence forces you to reread, it’s on me, not you.
- Evidence over enthusiasm – Excitement is fine; un‑checkable claims are not.
- Readers > algorithms – No SEO keyword soup. The headline says what the post delivers.
- Iterate in the open – Progress logs stay live so you can watch ideas evolve (or fail).
Thanks for stopping by. If the experiments help you think more clearly about AI—fantastic. If they help you catch me in an error—even better.
—Joe