Happy New Year

What 2025 Taught Me

Another belated one—and this time, it’s really belated. Happy Thai New Year again, I suppose. At this rate, Songkran is becoming my unofficial annual deadline for wrapping up the previous year’s thoughts. I’ll take it.

2025 was complicated. Not bad. Not great. Complicated. Same structure as last year: finance, business, and AI.

1. On Finance 📈

Last year I wrote: “I hope the trend continues into 2025 (spoiler alert: it doesn’t).”

I was wrong. Kind of.

Net worth grew 12.3%—almost identical to 2024’s 13%. On paper, another solid year. But the journey to get there was significantly bumpier. Trump’s tariff policies sent markets into convulsions early in the year. AAPL had a rough stretch as supply chain concerns and China headwinds created real uncertainty around one of my largest holdings.

For most of 2025, I genuinely thought I was heading into negative territory. The recovery was real but uneven, and a disproportionate chunk of it came in the final quarter. So: good outcome, deeply uncomfortable process. I’ll take the number—but I’m not going to pretend it felt good getting there.

The lesson I’m internalizing: concentration risk in individual stocks is psychologically expensive, even when it eventually pays off. VOO is boring. AAPL is volatile. Both are fine. But I need to be more honest with myself about what kind of investor I actually am versus the kind I imagine I am.

Since I’m already late, might as well write about tax. This year is rough. I paid ~$70k in income tax this year for selling AAPL down to 25% total portfolio. That’s a steep price to pay especially when my initial objective to sell was to buy a property in Thailand… A goal that hasn’t materialized yet.

Thing is, the selling was also to hedge against Forex volatility as well. 2024 saw $1 went from 36 THB down to 32… It’s a wake up call for me to make sure I keep enough in Thailand to avoid falling victim to currency fluctuation. It’s in nobody’s interest for THB to rise up that high: less visitors and less export revenue. But it still came to pass… what on Earth is happening?

The goal in 2026 is to trim down further AAPL concentration in my portfolio down to 10%. Keeping the cash for emergency use of up to 1 year worth of expense, and liquidate a bigger part of portfolio into cash. Rumor has it that Berkeshire Hathaway is holding a large amount of cash reserve for some reasons for quite awhile now. I’m probably not going to turn 31% to cash. But maybe 20% is a good call.

This year also marks the change in my expense calculation as well. Previously, I was following The 4% Rule (derived from the 1994 Bengen study and later the “Trinity Study”) and allocate more or less depending on how well my investment fares. I realize I’m not ready for the more unstable world of 2026 with that setup. So I cap my spending to 1.2m THB per year.

2. On Business 😐💼

2025 was the year I admitted something to myself: I’m not a product person. Not yet.

SaiJaiAI started with an ambition to build something real—a platform, a product with recurring revenue that doesn’t require me to personally show up for every engagement. That ambition is still there. But 2025 made clear that wanting to build a product is not the same as knowing what product to build.

2.1 The Thai market isn’t ready—or I’m not reading it right 🤷

The local AI market in Thailand is thin. Enterprises want bespoke solutions. SMEs want cheap (or free) tools they already recognize. Startups want to partner, not pay. There is a gap where a real AI product could live, but I haven’t found the right problem to anchor it to.

Is this a market-readiness problem? A distribution problem? A positioning problem? I genuinely don’t know. What I do know is that spending another year building something nobody is actively asking for isn’t a strategy—it’s denial dressed up as ambition.

2.2 Doubling down on services—deliberately, not by default 🔁

So SaiJaiAI is leaning harder into services. Not because I gave up on products, but because services keep the operation running, sharpen domain knowledge across different industries, and—more importantly—surface the actual problems people are willing to pay to solve. That last part is what I’m watching for.

The market is not ready. Instead of capitalizing on AI, companies are trying to standardize workflows and put down how things work into paper. If more service oriented means more potential product users in the future, then this is the way forward. Might as well speed up the AI infra adoption rate and make the $$$ along the way.

The goal isn’t to be a consultancy forever. The goal is to find the thing worth building by staying close enough to real problems to actually see them.

3. On AI 🤖🔧

2025 was the year AI demos stopped being demos. Last year I predicted agentic AI would be the next frontier. I was right. I just underestimated how quickly it would move from “interesting prototype” to “actually in production.”

3.1 Agents stopped being a buzzword 🧠

MCP (Model Context Protocol) went from an Anthropic blog post to a genuine adoption pipeline across tooling and enterprise integrations. Claude Code, Cursor, and similar tools became real daily-driver utilities—not just for hobbyists, but for working engineers. The jump in output quality was significant, but the more important improvement was in reliability. That’s the part that actually matters for production use.

The agents winning in the market aren’t the ones with the broadest capabilities. They’re the ones that fail gracefully and know when to ask for help. Turns out that’s also true of good consultants.

3.2 How did last year’s predictions hold up?

At the end of my 2024 recap, I called three things:

• Content aggregation and digestion — Partially true. AI-powered reading and summarization tools are everywhere now, but the raw ad-supported web isn’t dead yet. Legacy media is hurting, not gone.
• Personal data indexing — Slower than expected. The tools exist. Consumer adoption is still thin. Trust is the bottleneck.
• Zero-friction discovery — Hit or miss. Recommendation AI is meaningfully better, but most of it still optimizes for platform engagement, not your actual interests.

About one-for-three on full materialization. Not bad for a one-year-out call. I attribute it to market adoption enertia. I’m fully convinced the tech is now there. Just that people still prefer to do it the old way. Old habits die hard or something like that. But eventually, people will move forward to find other pastimes instead.

3.3 What I’m watching in 2026

2026 is already strange. Tariffs are reshaping global supply chains in ways that haven’t fully settled. Meanwhile, Anthropic and OpenAI are filing for IPO. What a strange world we live in. AI is accelerating at a pace that’s genuinely hard to price into anything. And I still don’t know what SaiJaiAI’s flagship product will be.

But I know more about what it won’t be—and that’s real progress.

Let’s see what the year brings.