I hope all of us and our families have a better year this year than last, however good 2025 was for you and yours.
The good news for 2025 includes
we are not in a major war
AFAICT no new major wars have started
our southern border appears to be under control
gas prices are down which will have run-on effects throughout the economy
the rate of inflation is holding fairly steady
Unfortunately, each piece of good news is accompanied by bad news
the war in Ukraine continues and shows little sign of ending
we are using military force far too frequently and casually
immigration enforcement has been harsh
electricity prices continue to rise which will also have run-on effects
asset prices are continuing to rise faster than inflation
One of the predictions for 2026 I’m running into fairly frequently is for a baby boomlet (possibly in 2027) among Swifties spurred by Taylor Swift becoming pregnant. Considering that I don’t frequent gossip sites or popular music sites running into that prediction so frequently is surprising.
Happy new year!
I’d be remiss to not make some predictions on 2026 (the comment is a response to the previous post and the next post).
1. The Ukraine war “ends” as in a long-term cessation in active hostilities. Similar wars of attrition (WW1 and US civil war) ended slightly after the 4th anniversary.
2. Democrats win the House of Representatives, but underperform expectations; and Republicans keep the Senate. The results leave both parties unhappy and leaves the 2028 field wide open as the first candidates publicly declare their Presidential candidacies by Christmas.
Democrats need to win less then 5 sits to win the House, but due to how districts were drawn in the 2022-2030 cycle, there aren’t very many “purple” seats to be won.
3. Inflation surprises to the downside in the first half (as the economy comes close to stalling), then the economy picks up and generates upward pressure on prices in the 2nd half.
4. There won’t be active hostilities in either Taiwan or Venezuela.
Now to some AI observations and predictions.
First, I spent the holidays using the latest models and tools — “Gemini CLI” and “Claude code”. Maybe I’m late, but I was impressed; those tools were able to generate applications or complete software engineering tasks that I estimate less then 3-4% of working programmers could successfully do. Eerily, the creator of Claude code said AI had written 100% of his contributions to Claude code in the past month, which brings up Anthropic’s CEO’s prediction that 90% of software written in Anthropic would be written by LLM’s 3 months ago.
Here’s my beliefs.
1. We should be cautious on being sure what’s the right solution to LLM’s causing job loss — like we should train software engineers to be “overseers” of agents. It makes 2 assumptions; we know the ceiling in the capabilities of models (and that ceiling is close to what models can achieve now), and forgetting ML is both about intelligence and learning. If AI can learn coding, why couldn’t AI be taught to be a better overseer then humans of other LLM’s.
2. If LLM’s keep improving at the rate they did in the last year, by the end of 2026 I think LLM’s will have made minor — but non trivial discoveries in mathematics or science.
3. The “moat” many companies think they have in their existing software intellectual property has been devalued. How companies react to the threat and opportunity that this creates will be a defining story of 2026.
4. In retrospect, it will be clear 2026 wasn’t when the AI bubble popped, but possibly when it began — the IPO’s of Anthropic / OpenAI in 2026 / 2027 will be signal of the mania phase of the bubble like the IPO of Netscape in 1995 or Amazon in 1997 did for the internet bubble.
CuriousOnlooker:
I agree with your prediction 2-4. I don’t know about #1. At this point I don’t think the war ends at all unless Ukraine collapses and/or Russia thinks it has won.
Regarding your AI predictions
1. They’re whistling past a graveyard.
2. That will be as much a liability as an advantage. Consider the viewpoint of non-tech company. Why buy now if next month’s/year’s product will be that much better?
3. Companies already overvalue old software. The biggest challenges are going to be supportability and legal liability.
4. I don’t know.