The Most Impactful 36 Months in Human History?
On Friday December 2, 2022, I received the following text from my brother: “I’m going to be out of a job.”
My brother, or Dr. Conklin as he wants me to refer to him at Thanksgiving dinner in front of mom, is a real estate professor at UGA. The tenure means it’s nearly impossible for him to lose his job.
Plagiarism scandal? The UGA real estate school shutting down? His horrific student reviews finally catching up to him?
No. The text included a link to something called ChatGPT.
I clicked on it and got an oddly simplistic search bar looking thing. I don’t remember what I typed first, but I remember it took about 3 questions before I asked something incredibly challenging - a pattern I saw repeat itself with everyone else I knew, too.
Simple. OK, that was decent. Little bit harder. OK, cool. Super hard. Whoa, mind blown.
I got nothing done for the next week. Research. Analysis. Stupid poems. I switched over to Wall-E for images. We were in San Diego at the time and I had it create an itinerary complete with dining options. Is this how the first people on the internet felt?
I sent it to everyone I knew. I wasn’t the only one. ChatGPT had over 1 million users within 5 days (and I was one of them!). FaceBook had taken 10 months. TikTok 9 months. ChatGPT was one weekend.
The genius was in the simplicity. No training. No implementation. No rules. Just type whatever popped into your head and let it do its thing.
Sam Altman is on every show espousing OpenAI’s mission to “benefit humanity” and that ChatGPT will forever remain a non-profit.
Today, ChatGPT has over 800 million active users. If ChatGPT was a country, it would be the third largest in the world. And that’s just one company. As I reviewed the past 36 months, what struck me most was the pace of change. There is a snowball rolling down a hill effect here, but I suspect that is my brain looking for a linear analogy that isn’t actually appropriate for exponential change.
Here's a recap of potentially the most impactful 36 months in human history.
2023
January - Microsoft invests $10B into ChatGPT
February - Google panic-releases Bard, which promptly provides the wrong answer on a live demo. Google stock drops 9% in one day and I learn what an AI hallucination is. “Does that mean some of the answers it’s been giving me might have been wrong?”
March - ChatGPT 4 launches and scores in the top 10% on the bar exam. Anthropic releases Claude, which strikes me as less fun but more accurate. The huge SREO abstract project I had done at the end of 2023 is largely worthless. What a waste of money.
April - Will Smith spaghetti video goes viral, with everyone suddenly feeling very good about their job security. I remember thinking, “ummmm, someone typed a few words into a text field and it made a video…are we sure this isn’t a game changer?”
May - ChatGPT hits 100mm active users, Hollywood writers go on strike. As adoption takes off, I learn what a “training window” is and which models have the most up to date info.
June - ChatGPT releases its first “agent”, which a tool that actually takes action. I try to have it book a vacation and it fails miserably.
July - NVIDIA, a total unknown to anyone outside hardware nerds, says chip demand has gone vertical. Market cap goes from $300B to over $1T.
August - ChatGPT launches an enterprise version, promising to not train its models on proprietary data. Sam Altman winks at the camera while making the announcement.
September - ChatGPT releases WALL-E 3, which improves videos but still can’t spell correctly in images.
October - Amazon, LinkedIn, and others release AI writing models. I’m now convinced my job is at risk, just like my brothers.
November - Sam Altman fired by the OpenAI Board. 745 out of 770 employees sign a letter saying they will quit if he’s not reinstated. Big money backers orchestrate Sam’s reinstatement and the Board is completely replaced. This in no way contributes to Altman’s God complex…
December - Google puts Bard out to pasture and releases Gemini, but again botches the release with an obviously staged demo. Apple says, “See? That’s why we don’t rush these things…”
2024
January - Apple leaks its coming Apple Intelligence (“AI” - get it?). I’m excited about the possibility of an AI model built right into my phone. Plus, it’s Apple so it must be awesome, right?
February - ChatGPT releases Sora, a mindblowing video tool. Hollywood freaks out, I make videos of fake property tours. My brother still has his job.
March - Anthropic releases Claude Opus, which instantly becomes my preferred model for hard stuff that requires precision. We ditch our AI abstract project after way too much money wasted. AI specialists don’t understand real estate nuances (cap rate vs rate cap or current balance vs original balance). We hire an AI Scientist and sit him next to our loan experts. Our kids are encouraged to attend state school.
April - Meta releases Llama, an opensourced LLM (read: free) simply to damage rivals with less cash than Zuckerberg. Grok is fun for images of real people. Real estate people hate tech but love the idea of data centers as an asset type.
May - ChatGPT releases a multimodal model. I can switch seamlessly between typing and talking, images and text, etc. I use voice to coach me through a chess game against a chess app.
June - Apple officially (and finally) releases Apple Intelligence (“AI” - get it?) and I immediately rush to try it. It sucks so bad I assumed I was using it wrong. It’s somehow as bad as Copilot.
July - Llama 3.1 updated context window allows for nearly 3x the capacity of what the original ChatGPT provided. What took OpenAI billions in compute and proprietary data, Meta gave away for free 20 months later. It dawns on me that “first mover advantage” might not apply to AI…
August - ChatGPT hits 200mm active users. I have over 20 meetings on a west coast trip where every single company still downplays AI as a fad. Not some. Not most. Every single one. I was floored.
September - ChatGPT gets cheeky with names and releases “strawberry”, the first reasoning model. We trade speed for accuracy and it spends more time thinking. It successfully completes PhD level tasks in physics. My brother’s job might be in jeopardy after all.
October - ChatGPT raises $6.6B at a $157B valuation while releasing “Search”, very blatantly going after Google. At this moment I realize that ChatGPT is eventually going to have ads just like Google so those investors can turn a profit. Gross.
November - AI fake videos and images circulate, raising concerns about election interference. Meanwhile, every model drives me nuts by ignoring simple instructions about not using em dashes.
December - OpenAI releases the full o1 and o1-pro. The "thinking" time drops, and accuracy on math benchmarks hits 90%+. NVIDIA's AI workhorses chips dropped from $40K to $20K as supply caught up with demand.
2025
January - China’s DeepSeek releases R1, an open-source reasoning model that rivals o1, causing a mild geopolitical panic. China spends $500B on this model only to give it away for free and pinky swear promises not to give your data to the Chinese government.
February - OpenAI drops GPT-4.5, a massive scale-up of GPT-4, described as their "last non-reasoning model." It feels human.
March - Google strikes back with Gemini 2.5, the first Google model I actually enjoy using. Google promptly ruins any good will by trying to analyze every cell in my Google sheets. Claude releases Code, and our engineering team’s capacity jumps 30%. I wonder if UI is dead? ChatGPT raises $40B on a $300B valuation.
April - ChatGPT releases 5.0 so we don’t have to choose the model anymore, it does that for us. Zuckerberg responds with Llama 4, built on a supercluster of 350,000 of the most powerful NVIDIA chips to power this free model. It requires the same amount of electricity to run as the entire city of Santa Fe. Zuckerberg is giving away a $10.B tool just to damage ChatGPT and Gemini.
May - Google releases “AI mode” and all that time we spent on SEO is rendered worthless. Copilot is still garbage.
June - NVIDIA announced the Blackwell chip had “reached full-scale production,” and that in one quarter alone Blackwell-generated revenue hit US$11 billion, marking the fastest product ramp in company history. Market cap exceeds $3 trillion. It is becoming apparent to even me that AI will require insane infrastructure support.
July - ChatGPT Agents launches and this time actually does book my vacation. Unfortunately, it also kills the 4.5 APIs which many companies had started leveraging. This reminds everyone that we are subject to the whims of the few companies controlling the AI models.
August - another west coast trip with over 20 meetings and every single one wants to talk AI. Not some. Not most. Every single one. I was floored. I watched as Waymo self-driving taxis behaved like humans, both good and bad, in downtown LA.
September - Gemini becomes my preferred model. It is more accurate and faster then any other model, a great combo of ChatGPT creativity and Claude accuracy. It seamlessly integrates images into a conversation. I’m pretty sure Gemini is trying to drive ChatGPT out of business by providing all of this for nearly free.
November - As we hit the 3-year mark, ChatGPT approaches 800 million weekly users. OpenAI releases GPT-5.1 ("personalities" update). We have moved from a funny poem bot to a global intelligence utility in exactly 1,096 days.
December - NVIDIA market cap of $4T exceeds the GDP of every nation in the world except the US and China. ChatGPT is worth $500B, more than Coca-Cola and Disney combined. Electricity demand is going to be a problem. Without ads, ChatGPT will struggle to compete with Google and Meta, who have ad revenue streams to fund operations.
It's been a whirlwind 36 months.
I’m obviously an AI fanboy and people ask me if AI is a bubble. I don’t know. I can’t tell if these valuations are insanity or reasonable. Valuations might collapse, but I am certain AI isn’t going anywhere. Amara’s Law says we overestimate the impact of new technology in the short run, but underestimate it in the long run.
AI will have a bigger impact on our lives than the internet did, I am certain of it. The only parallel I can think of is the advent of electricity, which is ironic given AI’s dependency on it.
But AI is the first invention that can, itself, invent. A tool that can create other tools. There is a spiraling effect that I don’t fully understand, but I do know it is driven by actual demand. And that means it’s not going anywhere.
The last 36 months might go down as the most impactful in history. An argument can be made for other time periods like WWII, but absent extinction level events I’m not sure there’s a true parallel. Everything, everything, is changing.
My brother still has his job and I still have mine…for now…but even in the face of that uncertainty, I can’t wait to see what the next 36 months will bring. I just hope it's not an extinction level event...
