Fractional CTO Tech Stack: What Worked For Me in 2025

Quick reminder that today at 11am I will be in conversation with John Durrant of Human-Centric engineering discussing everything from AI to fasting. Join us!

---

Following in the spirit of learning and sharing knowledge from yesterday’s post, I thought I’d share some of the tools and tricks that have worked for me as a fractional CTO in 2025. If you’re curious what a fractional CTO is and what one does, check out these first.

What does a CTO actually do?

What does a fractional CTO do?

The following article was written largely in collaboration with AI on a walkabout because, well, it felt appropriate :)

Let's walk and talk

As a result, expect fewer typos, and also a bit less of the questionable grammar flair I usually bring to you.

Note - I’m not paid for promoting any of these products, but I did include a referral link for WhisprFlow because what’s the point of having a newsletter if you can’t at least bank some free credits?

---

Here’s the thing about being a fractional CTO in 2025: your tech stack is your competitive advantage.

Not in the “look at all my shiny tools” way. In the “I can move 3x faster than I could last year” way.

This year has been transformative. The AI tools that were clunky experiments in 2024 are now genuine productivity multipliers. But here’s what nobody tells you: the winning strategy isn’t using one AI for everything. It’s about knowing which tool excels at what.

This article? A perfect example. I’m speaking these thoughts into VoiceNotes, and an LLM will polish them into what you’re reading now. That’s the new workflow.

Let me walk you through what’s actually working.

Voice: The New Keyboard

I’ve fundamentally changed how I create content.

WhisprFlow is my go-to for any text input across mobile and desktop. Quick emails. Slack messages. Document edits. It’s faster than typing and surprisingly accurate.

But for deeper thinking? That’s where Voicenotes comes in.

I’ll dump entire article concepts into it. Strategic thoughts about client problems. Complex technical explanations I’m working through. The app captures it all, usually as I’m out for a walk, and then I let an LLM transform the raw transcript into structured content.

This isn’t just about speed. It’s about removing friction from the creative process. When you can think out loud without worrying about sentence structure, you explore ideas differently. You go deeper. You also get some more steps in!

The LLM Landscape: Why I Use Three Different Models

Here’s where it gets interesting.

I started 2025 using ChatGPT for everything. Business proposals. LinkedIn posts. Client reports. It worked fine, especially since it had accumulated rich context about my business over time.

But then I discovered something important: different LLMs have different strengths.

Claude: My Writing Partner

For anything that needs to sound like me, I use Claude. Especially when I’m converting Substack articles into LinkedIn content. It captures my writing style better than any other model I’ve tried.

There’s something about how Claude structures sentences and maintains voice that just clicks. When I get Claude’s output, I rarely need to heavily edit it. It already sounds like something I’d write.

Gemini: The Enterprise Workhorse

One of my clients uses Google Workspace exclusively. For their work, I’ve gone all-in on Gemini.

Why? Enterprise-grade security around data and training. When you’re handling confidential company information, that matters. A lot.

But here’s the surprise: Gemini might be my most-used model now.

Not because it’s “better” than the others in some abstract sense. Because it’s integrated. It’s built into Docs. Into Gmail. Into the entire Google Workspace suite.

The latest Gemini 3 release really levelled up the experience. The integration keeps getting better. And when a tool is already where you’re working, you use it more. It’s that simple.

ChatGPT: The Business Memory

I still use ChatGPT for specific business tasks. Proposals. Financial modeling. Strategic planning. I also use it for personal stuff, like inventing recipes or exploring travel options.

Why keep using it? Because it has the most accumulated context about my business. That historical knowledge is valuable. It understands my business model, my clients, my goals. It’s also still the most ‘consumer friendly’ app for casual stuff.

But its dominance is fading. As the other models learn my business, I find myself reaching for ChatGPT less and less, at least in a professional context.

Coding: Back to the Terminal

For the small amount of hands-on coding work I’ve been doing, I’m leaning heavily on Claude Code.

Working directly in the terminal feels right. It’s fast. It’s focused. It doesn’t try to do too much.

I’ve been watching the agentic coding space closely though. Antigravity is on my list to try next year for rapidly spinning up front-end focused prototypes. The promise of AI that can scaffold entire applications is compelling.

We’ll see if it lives up to the hype.

The Creative Side: Images (But Not Video Yet)

Image generation has become genuinely useful.

I’ve been experimenting with it for one client’s marketing concepts. The latest Nano Banana Pro (Google’s Gemini 3 Pro Image model) has been particularly impressive. The text rendering is sharp, the reasoning behind image composition is solid, and you can iterate through dozens of concepts in the time it used to take to brief a designer on one.

What makes Nano Banana Pro stand out is its ability to generate accurate text directly in images and its connection to Google Search for real-world context. Perfect for creating marketing materials, infographics, and client presentations.

A fun Christmas marketing POC I did for my client Towards

But video? Still not there.

I’ve tried the various video generation tools. They’re impressive as tech demos. They’re not yet useful for actual client work. Maybe 2026.

Meeting Management: The Unsung Hero Tools

Two tools have transformed how I handle meetings:

Gemini for transcription on my Google Workspace client’s calls. The integration is seamless. It just works.

Granola for everything related directly to my business. Other clients. Partnership discussions. Business development calls.

Having accurate transcripts and summaries changes everything. No more frantic note-taking during important conversations. No more “wait, what did they say about that deadline?” moments.

You can be present in the conversation because the tool is handling the memory.

What’s Still Missing

Here’s the honest part: I’m probably operating at 60% of what’s possible.

There’s so much more I could be doing with automation and integration. My CRM situation is embarrassing. I’m tracking client contacts and interactions far too manually. There are definitely better ways to manage this with AI assistance.

I experimented with MCP (Model Context Protocol) integrations with Claude earlier this year. The potential was obvious. The execution was janky and unreliable.

I dropped it. But I’ll revisit in 2026. These things need time to mature.

The Meta-Lesson

The biggest thing I’ve learned? The best AI stack is the one that adapts to your workflow, not the other way around.

I’m not using the “best” AI for every task. I’m using the right AI for each context:

• Claude when I need my voice

• Gemini when I’m in Google Workspace

• ChatGPT when I need business memory

• Voice tools when I need to think out loud

• Specialized tools when I need specialized results

That’s the actual winning strategy.

Your stack will look different. That’s fine. That’s good, even. The question isn’t “what tools does everyone else use?” It’s “what tools let me do my best work?”

Over to You

This is my 2025 stack. It’s evolved significantly from where I started the year. It’ll probably look different again by this time next year.

What am I missing? What tools are you using that are genuine game-changers?

I’m especially interested in hearing about:

• Better CRM/contact management approaches

• Automation workflows that actually work

• Tools for spinning up prototypes faster

• Anything that’s working for you in the fractional/consulting space

Drop a comment or reach out. Let’s figure out what 2026’s stack looks like together.

---

This article was written using the exact workflow I described: spoken into Voicenotes, structured and refined by Claude, edited by a human (me). Fun, huh?