From Chess Bots to ChatGPT: How AI Went From Geek News to Daily Dependency
AlphaGo beating some Go grandmaster was cool. But it didn’t help you ship that feature before Friday.
Then one random November evening, a chat box appeared on a website, and suddenly your PM, your cousin, and your dentist all had opinions about AI.
Table of Content
- When AI was just a headline
- Pre-ChatGPT: AI as infrastructure, not coworker
- November 30, 2022: the chat box moment
- Why this one actually felt different
- For developers: from demo toy to daily tool
- The paradox: AI is old, adoption is new
- The new baseline: AI in the loop
When AI was just a headline
For years, “AI” lived in the tech news section you read while pretending to work.
Back in 1997, IBM’s Deep Blue beat Garry Kasparov in a six-game chess match, becoming the first system to defeat a reigning world champion under standard tournament conditions. Twenty years later, people still write think pieces about how that match “changed everything,” which is technically true, except for your daily standup.
Fast-forward to 2016: DeepMind’s AlphaGo walks into the ancient game of Go, looks at centuries of human intuition, shrugs, and beats Lee Sedol 4-1. Again, huge milestone, tons of headlines, lots of “AI vs humanity” drama… and then on Monday you still had to write your own pagination logic.
AI was impressive. But it was over there, in research labs, documentaries, and conference keynotes. Not in your terminal.
Pre-ChatGPT: AI as infrastructure, not coworker
Before ChatGPT 3.5, AI was already everywhere. Just invisible.
It picked what you saw on social media, routed your emails into “Spam” or “Promotions,” scored your credit risk, and helped ad platforms follow you around like a clingy ex. From a dev perspective, it looked like this:
- A black-box API you call for “recommendations.”
- A model that someone in “data science” trains once a quarter.
- A slide in a pitch deck with a cloud icon and the word “AI” slapped on it.
You knew it was there, but it wasn’t with you. It didn’t sit next to your code editor. It didn’t argue with you about whether this function name is garbage.
The mental model was: AI = huge compute, PhD, cluster, weeks of training, weird Jupyter screenshots. Cool, but not something you spin up between two bugfixes.
November 30, 2022: the chat box moment
Then came ChatGPT 3.5.
On November 30, 2022, OpenAI dropped a simple web UI: a text box and a button, branded as a “research preview” powered by GPT-3.5. Under the hood, sure, lots of fancy language modeling. On the surface? It looked like Slack but with fewer reaction emojis.
It reached a million users in five days and tens of millions more in the following weeks. Translation: one day you were mocking AI hype; the next day your non-technical friend was sending you “look what it wrote for my marketing email” screenshots.
The big innovation wasn’t just the model. It was the UX:
- No setup.
- No API key.
- No reading a 30-page doc about “temperature” and “top-p.”
Type words. Get words back. That’s it. And that was enough to flip AI from “some lab thing” to “I’ll just ask the bot.”
Why this one actually felt different
Deep Blue and AlphaGo showed that machines can beat humans at well-defined games with clear rules. Impressive, but also extremely narrow: no one uses a Go bot to write an email to HR. (Yet.)
ChatGPT 3.5 did something else: it pretended to understand everything you throw at it, in natural language.
You could:
- Paste a horrible legacy function and say “explain this like I’m on my fifth coffee.”
- Ask for “a bash one-liner to do X, please don’t use
xargsI hate it.” - Draft emails, blog posts, commit messages, migration plans, or that Jira ticket that should have been closed three sprints ago.
Same underlying idea (statistical modeling of text), but the interface made it feel like a coworker who never sleeps and sometimes hallucinates. And that’s when AI stopped being background infrastructure and started being foreground workflow.
For developers: from demo toy to daily tool
If you write code for a living, the shift was brutal and fast.
Day 1: “Lol, I asked it to write FizzBuzz, look at this.”
Day 3: “Ok, it wrote a decent unit test. That’s… useful.”
Day 10: “Why am I still manually writing boilerplate? What am I doing with my life?”
Suddenly, the typical dev day quietly changed:
- You still Google, but now you also paste stack traces into a chat.
- You still read docs, but only after the bot gives you a rough mental model.
- You still write code, but the boring 80 percent is now “generate, tweak, test” instead of “stare at blank file.”
The magic wasn’t that it knows all the answers. It’s that it gives you a first draft of almost anything:
- SQL queries.
- API specs.
- Refactoring suggestions.
- Regex that you swear you’re going to understand this time.
And yes, sometimes it’s wrong. But so is Stack Overflow, your memory, and that random GitHub Gist from 2015. You still review everything.
The paradox: AI is old, adoption is new
Here’s the funny part: the term “artificial intelligence” is older than many devs currently arguing about it on Twitter. We’ve had expert systems, neural nets, Bayesian stuff, you name it, for decades.
What changed with GPT-3.5 wasn’t just capability. It was:
- Packaging: a browser tab instead of a research paper.
- Access: free (at first) and frictionless.
- Use cases: not “beat grandmaster at game,” but “help me not look stupid in front of my team.”
AI went from “something companies have” to “something I use” almost overnight. From infrastructure to interface. From backend service to pair programmer.
That’s why it feels like it “arrived suddenly,” even though the underlying research had been grinding forward for years.
The new baseline: AI in the loop
We’re now in a weird, hybrid world:
- You still need to understand what good code looks like.
- You still need to know architecture, trade-offs, and security basics.
- But you also have this autocomplete-on-steroids sitting next to you, ready to generate an entire microservice if you let it.
The dev skill set quietly upgraded:
- Knowing how to ask becomes as important as knowing what to type.
- Checking AI output becomes a core competency, like code review or testing.
- Your “I don’t have time” excuse starts to sound weaker when a bot can draft your first version in 30 seconds.
The same way we stopped manually managing memory everywhere, we’ll probably stop manually writing certain categories of code. The job doesn’t disappear; the baseline just moves.
We went from watching AI win at chess and Go on TV to having it sit in our browser, inside our IDE, and eventually in our terminal history. And that, more than any fancy model architecture, is what made it feel like AI suddenly took a giant leap into our lives.