Hey everyone! I hope you had a great holiday season and you are ready to kick off 2026 in style. Before we turn the page, I wanted to close out the year with a quick reset on what I learned in 2025 and a simple system to start the new year strong.

🧠 Founder’s Corner: 3 lessons from my 2025 AI journey, from embracing discomfort to treating AI like a system and recognizing how early we are.
🎓 AI Education: a high-level recap video that ties the full LLM 101 series together so it actually sticks.
✅ 10-Minute Win: a Financial Resolution Action Plan and accountability tracker you can run in one sitting and reuse every week.

Let’s dive in.

1) One in a million: celebrating the customers shaping AI’s future

Summary: OpenAI says over 1 million customers use its products globally, rolling out ChatGPT for writing, coding, research, analysis, and building agents; 75% of customers said AI helped them complete tasks they couldn’t do before.

Why it matters: This is adoption going from “pilot” to platform. More customers → more workflows → more lock-in → more enterprise spend.

2) How Oracle became a poster child for AI bubble fears

Summary: Oracle has become a market proxy for “AI bubble” anxiety: massive AI-infra ambition, heavy capex/debt questions, and investor sensitivity to any sign that AI profits will take longer than promised.

Why it matters: If the AI trade gets a reality check, it’ll likely start with capex + financing + timelines—not model demos.

3) Nvidia aims to begin H200 chip shipments to China by mid-February, sources say

Summary: Nvidia plans to restart China shipments of H200 AI chips around mid-February 2026, using existing inventory (roughly 40,000–80,000 chips equivalent), pending approvals.

Why it matters: Export policy → compute access → model capability. This is one of the biggest levers in the global AI race.

4) Illinois climate groups want pause on data centers

Summary: Illinois environmental groups are pushing for a pause on new data centers, warning the AI/crypto boom could strain energy supply, raise prices, and increase pollution/water stress. 

Why it matters: AI is colliding with local power politics. Permits, grid constraints, and backlash can slow “AI everywhere” more than model progress does.

5) Google Cloud CEO reveals 10-year AI strategy on power and silicon

Summary: A recap of Thomas Kurian’s comments: Google has been executing a decade-long plan built around two bottlenecks—specialized silicon (TPUs) and power availability/efficiency—not reacting to a sudden trend.

Why it matters: The real AI moat is shifting to energy economics + compute efficiency. Whoever can scale capacity without getting grid-blocked wins.

2025 was the year the pace of AI change became impossible to ignore. Whether it’s stock market bubble talk, new model breakthroughs, agentic tools, or the infrastructure spending behind it all, one thing is clear: the pace of change is accelerating, and it feels like we are just getting started. For me, 2025 was the year AI went from interesting to unavoidable, and it pushed me to learn in public by building content that strengthened my own foundation while helping others start theirs.

Since launching in September, this community has grown to over 100 subscribers, and I’m genuinely grateful for the support along the way. Over the last few days, I’ve been reflecting on the journey, celebrating the wins and being honest about the challenges. The lessons I learned in 2025 gave me the foundation to sharpen what I’m building and show up with more confidence in 2026. I want to share these lessons to help you build your own confidence, stay curious, and set yourself up for what’s coming in 2026.

Lesson 1: Discomfort is the tuition for AI fluency

I had no idea how AI actually worked when I started to get serious in late 2024. Most of the concepts were foreign, and like a lot of people, I was only using AI for fun. It was a basic chat here and there, a funny AI-generated photo for the group chat, or a question that could have been answered with a normal Google search. I had to adapt. If I wanted to get better, I had to get uncomfortable.

For me, that meant learning unfamiliar topics, trying tools that felt intimidating, and being willing to look a little clueless while I built in public. Only through reflection have I realized how important this was for building AI skills and knowledge. Every error or break forced me back into AI to troubleshoot and solve the problem. Each experiment with a new platform built skills I never thought I could develop. Every successful launch built my confidence and helped me find my voice.

Being uncomfortable is hard, but it is also where the learning happens. Acknowledge the discomfort, then set small goals that let you move at your own pace. One rep a week is better than sitting on the sidelines. I strongly believe AI will become a bigger part of everyday life, especially in the workplace. The people who lean into that challenge now will be set up for success in 2026 and beyond. 

Lesson 2: AI is not magic, it is a system 

It’s easy to be blown away by AI tools, no matter your level of expertise. Innovation is happening daily, and we get to benefit from the competition between the big labs. It’s exciting, but the wrong approach can build bad habits and quietly lower the quality of your outputs. I fell into the same trap a lot of people fall into when they first start using AI. I honestly thought magic was happening behind the scenes. I assumed any prompt would produce the output I had in my head. Only through practice, learning, and a few slices of humble pie did I start to realize these tools are systems. Like any other system, there are rules and constraints you need to follow to get high-quality output.

That realization pushed me to learn how LLMs actually work, so I could create better content for the newsletter. I built a simple system that treated AI like a partner by giving it clear instructions, goals, and context. Once you put this into practice, your outputs get better fast, and you start building real confidence in how you use AI. 

Lesson 3: We are early, but change is coming 

I spend a lot of time absorbing information to stay current with the trends shaping the AI world. I pull data points and opinions from a mix of places, including podcasts, news, and thought-provoking conversations with friends and colleagues. I was genuinely shocked when my Spotify Wrapped said I listened to more than 46,000 minutes of podcasts in 2025. I’d estimate at least half of that was directly related to AI or AI-adjacent topics. The more I learn, the more I believe the real impact of AI is still ahead, and that 2026 will lay the foundation for a very different world.

On the surface, there is plenty of evidence that suggests AI is already in a mature state. The big labs are pushing the boundaries of what is possible, and every new model release seems to break benchmarks set by the last one. So why do I feel strongly enough to say we are just getting started? The answer is the gap between consumer adoption and enterprise adoption. At home, you can discover a new tool and start using it the same day. At work, you are limited to whatever is approved and rolled out. It’s also easier to organize your own data and build personal workflows than it is for a large organization to modernize systems, clean up data, and wire everything together in a way that can support agentic workflows at scale. We are still early in this shift, even if the technology itself is moving fast.

Gartner predicts that 40% of enterprise applications will include task-specific AI agents by the end of 2026, up from less than 5% in 2025. If that happens, AI will stop feeling like an “initiative” and start feeling like the default way work gets done. That is why it feels imperative to take action now. Start simple by picking the biggest problem inside your role and working backward from the outcome you want. Look for the workflow underneath it, then use data to define what good looks like and where automation can actually help. Push for outcomes that deliver ROI, but also build trust and drive adoption responsibly. That mindset is how you stay ahead of the curve and become the person who helps your team navigate what comes next.

2025 taught me that AI rewards the people who stay curious and keep taking reps, even when it feels uncomfortable. The technology is moving fast, but the bigger shift is how quickly it is being woven into everyday life and the workplace. My goal going into 2026 is simple: keep learning in public, keep refining my systems, and keep sharing what works and what breaks along the way. If you take one thing from these lessons, let it be this: start small, stay consistent, and build the habits now that your future self will thank you for.

Over the past few weeks, you have read through our four-part series on Large Language Models. We started by establishing a clear mental model of what an LLM actually is beyond the marketing answers. We looked at the training phase where the model learns patterns from huge amounts of text. We examined the distinct "use phase" where your specific prompts are turned into tokens, and we finished by discussing the critical limits you need to know, such as hallucinations and context windows.

That is a lot of information to process. To help you digest the full series, this video will serve as a high-level recap. Our goal is to ensure that by the end of this overview, you truly understand what is happening under the hood when you ask AI to do a task. Let’s put a clean frame around the building blocks you have picked up so far.

✅ Financial Resolution Action Plan + Accountability Tracker

A financial resolution without a system becomes a January feeling you forget by February. The problem isn’t motivation, it’s clarity + follow-through. In 10 minutes, you’ll use ChatGPT to turn one money goal into a concrete weekly plan, then paste an AI-generated tracker into Google Sheets so you can stay accountable with a 2-minute weekly check-in.

Model tip (optional): If you see GPT-5.2 Thinking (or a “Thinking” option) in your ChatGPT menu, use it for Step 2–4. If you don’t, the workflow still works with the default model.

Step 1 — Choose ONE resolution and define success (2 minutes)

Pick one goal you can measure in dollars and a date. Examples:

• “Save $1,000 by March 31”
• “Pay off $2,500 of credit card debt by June 30”
• “Invest $50/week for 12 weeks”

Write 3 lines (rough is fine):

• Goal: (dollars + date)
• Why it matters: (1 sentence)
• Your constraint: (what makes this hard: irregular income, holidays, debt, etc.)

This keeps ChatGPT focused and keeps you honest.

Step 2 — Have ChatGPT build your action plan + guardrails (3 minutes)

Paste this prompt into ChatGPT:

Role: You are my financial resolution coach.Goal: Create a simple, realistic action plan I can follow weekly.

My resolution:

• Goal (dollars + date): ___
• Why it matters (1 sentence): ___
• Constraint (what makes it hard): ___

Tasks:

1. Rewrite my goal as a SMART goal (specific, measurable, time-bound).
2. Break it into weekly targets (12 weeks max).
3. Give me a Weekly Playbook with 3–5 actions I repeat (example: “move money on payday,” “one spending rule,” “one income action”).
4. Create 5 If–Then rules to prevent failure (example: “If I overspend this weekend, then I do X on Monday to recover”).
5. Output a short “Minimum Viable Week” version (what I do if I’m busy/stressed).

Rules: Keep it beginner-friendly. No investing or debt advice beyond planning behaviors.

You now have a plan that survives real life.

Step 3 — Have ChatGPT generate the tracker table to paste into Sheets (3 minutes)

Now paste this prompt:

Create an Accountability Tracker I can paste into Google Sheets.

Requirements:

• Output CSV only (no commentary).
• Row 1 = headers.
• Include columns: Week Start Date, Weekly Target ($), Actual Saved/Extra Paid ($), Delta ($), Cumulative Target ($), Cumulative Actual ($), Status (On Track / Behind / Ahead), Notes
• Put formulas in the right cells (assume the first data row is row 2):
  • Delta = Actual - Weekly Target
  • Cumulative Target = sum of Weekly Target to date
  • Cumulative Actual = sum of Actual to date
  • Status = On Track if Cumulative Actual >= Cumulative Target, otherwise Behind (Ahead if it’s meaningfully above)
• Pre-fill Week 1 with sample values and formulas so I can see it working.
• Use weekly targets from the plan you created above.

Then:

1. Copy the CSV output
2. Open a blank Google Sheet → click A1 → paste
3. Delete the sample “Actual” number and replace with your real weekly updates.

Step 4 — Run a 2-minute weekly check-in (2 minutes)

Every week (same day), copy the last 1–3 rows from your Sheet and paste into ChatGPT with this:

Weekly Check-In: Here are my last updates (table rows): [PASTE 1–3 ROWS]

Do this:

1. Tell me if I’m On Track / Behind / Ahead (one sentence).
2. Give me one adjustment for next week (not five).
3. Rewrite next week’s 3–5 actions as a simple checklist I can follow.
4. If I’m behind, propose a recovery move that doesn’t require extreme sacrifice.

That’s your accountability loop: quick, non-dramatic, effective.

The Payoff

You end with a real system: a SMART goal, weekly targets, guardrails for bad weeks, and a tracker that makes progress visible. The magic isn’t the spreadsheet, it’s the weekly feedback loop. Do this for 8–12 weeks and you stop “hoping” you’ll be different—you become consistent.

Transparency & Notes for Readers

• All tools are free: ChatGPT + Google Sheets.
• Limits: ChatGPT Free has usage limits; if you hit them, run the weekly check-in later or keep it shorter.
• Accuracy: This is planning and accountability, not personalized financial advice or a replacement for a professional.
• Educational workflow — not financial advice.