Howdy wizards,
Here’s what’s brewing in AI.
The 8 AI tools I actually use (mid 2026 update)
In the last months, I’ve focused exclusively on how companies are implementing AI and the AI news landscape. I haven’t been sharing much about my own experience in building with AI.
That’s because I value the outside perspective, learning about how other people are using AI is fascinating, and also because I have a lot of unique data from Context Windows that I think the world needs to see.
That said, I’d like to get back to also sharing a bit more of how I personally use AI over the next weeks. Let’s start with an updated overview of the tools I actually use from day-to-day.
So that you understand my background better, these are the main purposes I use AI for:
building & maintaining contextwindows.ai
researching and writing this newsletter
consulting work for a tech company
learning things in my daily life
My toolset is very simple, and I like it that way. I use few tools, so that I can truly become proficient in them. As often as possible I try to use bare metal — tools that are directly interacting with a language model (e.g. Claude, ChatGPT), or in a direct way make it more effective for me to do so (e.g. a transcription app).
I find having a small set of tools that are highly adaptable to any task is simpler, cheaper and more educational for me than buying pre-packaged workflows. Thought there are exceptions.
It’s important to understand that the tools I choose to use may differ radically from what you should be using. It all depends on your domain, technical ability, interest in learning, free time to get good at new tools, etc.
Case in point: I don't make video, so there's no video tool anywhere in my setup. But if that's part of your work, it's exactly the kind of thing your list should have, and today's sponsor is definitely worth a look.
IN PARTNERSHIP WITH VIDU
Most creators struggle to maintain consistency across shots, wasting time and budget on unpredictable AI video tools.
Vidu brings structure and creative control to high-volume workflows. It transforms text, images, and visual references into high-quality, visually expressive videos in seconds.
Reference-to-video: Maintain character, object, and scene consistency across cuts.
Rapid generation: Create cinematic or anime clips in just 10 seconds.
Cost-efficiency: Scalable production with pricing as low as $0.015/sec.
From action-heavy anime scenes to dialogue-driven storytelling, Vidu helps you eliminate random drafts and ship multi-shot stories at a fraction of traditional costs.
PS the same video generation is also available via API, free to test.
❦
Without further ado, these are the 8 tools I’m using right now:
AI tools:
Claude Code ($100/mo) with a Max subscription (I’m currently on the 5x plan)
Paraspeech ($10/mo), an awesome transcription app that runs locally on your computer. Because typing text to AI is so 2025 — using your voice to type means saving time, freeing yourself from the screen (I often walk around the room talking), and getting better outputs because you actually bother to articulate things in detail.
Granola (free), a meeting note taker. I use it for any calls I have during the week. It’s insane not to. You literally take the burden of having to remember every minute detail from conversations off your shoulders. And you can Q&A the notes at any time. What I love about Granola specifically is that it sits on my computer, in the background, outside of the meeting room. And overall a delight to use.
CatchAll ($50/mo), a deep web search API. LLMs’ native web search is very limited; CatchAll digs up everything related to your search. And it uses AI to validate, deduplicate and enrich your results. This allows me to create detailed datasets on any event I want, that my agents can readily use. I use it to discover new published case studies about AI implementations, which then hit my enrichment pipeline, and then get featured on Context Windows.
The Gemini API. Gemini Flash and Pro power some of the enrichment I do of case studies on Context Windows. I tested the leading models for this task and Gemini was the best tradeoff in terms of performance and price.
Infrastructure tools I use to build with AI:
Hetzner VPS ($5/mo). Think of a VPS like your own little computer sitting in a datacenter somewhere that you rent and that’s always on. It costs ≈$5/mo and the possibilities when combined with AI are endless. I’ve installed Claude Code on the VPS with its own Claude Pro ($20/mo) subscription. That Claude agent has a bunch of daily tasks I invoke as cron jobs (scheduled tasks) and it just does them without me. It has access to GitHub so a lot of it is maintenance tasks on my existing projects, as well as data scraping. I also like that it’s LLM agnostic — if I ever want to switch from Claude to another model, I can easily switch and all my workflows would still work.
Supabase ($20/mo), a Postgres database that talks well with AI. This is the base of any product I’m building. If you’re trying to build things but don’t have a connection to a database where you can store things, you’re not going to get very far. Importantly, I use the Supabase MCP so that the agent can always read data and make changes to the database without me — I almost never go into the Supabase platform myself.
Vercel ($20/mo), a hosting platform. If you want to put your apps out into the world and not only on localhost:3000 then you need a way to host it. Vercel gives me everything related to hosting and is easy to use.
That’s it.
Remember, the important thing isn’t which tool you use, as long as you get the job done. You don’t need to try or use every shiny new thing that comes out. Choose deliberately and switch out tools when you need to. I add or switch around 1-2 tools in this stack per year.
Very good at a handful of tools beats half-decent with a lot of different ones.
❦
You are a delight.
Dario
What's your verdict on today's email?
See which AI use cases are paying off with Context Windows Pro
Most companies pick AI use cases by brainstorming internally. 90% of those initiatives fail.
I’ve created Context Windows just so you can pick the winners.
🟦 Find high-performing use cases from 2,000+ companies at contextwindows.ai, or book a demo with me
Disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes work with sponsors and may earn a commission if you buy something through a link in here. If you choose to click, subscribe, or buy through any of them, THANK YOU – it will make it possible for me to continue to do this.