It’s now as hard to imagine working without AI tools as it once was to imagine life before Google Search. Here’s my toolkit in early 2026: what I pay for, what I rely on daily and why.
ChatGPT
I’m locked in at US$22/month, and I get good value from it. The accumulated knowledge I’ve built up over time makes switching costly. I use it for:
-
Suggesting camera kit to take on a trip (too many decisions)
-
Writing Applescripts and Perl scripts
-
Optimising my solar and battery at home
-
Fact checking
-
Summarising lengthy podcasts
-
Creating fun images and cartoons
Indispensable.
Atlas Browser
Tied into ChatGPT, but I mainly use Chrome. A bit of needless overhead for simple web surfing sessions, although it can track and analyse my browsing behaviour and remind me where my time sinks are.
Claude.ai
My go-to when it’s the better tool for the job:
-
Parsing PDFs into useful data
-
Charting economic data
-
Getting a second opinion on something where I’m receiving “editor” feedback
AI/LLM tools continue to evolve rapidly, so it’s good to have a few to call on.
Perplexity.ai
For search. I use it multiple times per day but not enough (yet) to pay for it.
Google Gemini
For querying YouTube videos. Native access to transcripts makes it a killer application. For example: “Summarise this video…” or “The video says ‘5 reasons why…’ so tell me what those reasons are.”
Notebook LM by Google
Great for loading your own specific sources and directing queries at them:
-
I have our Local Planning Ordinance (rules for land use and development) loaded as a PDF. It’s 1,300 pages long. A normal keyword search finds a lot of low-value results. Notebook LM provides smart search and retrieval of relevant information.
-
Good for loading technical documents (like a user manual for a car or camera) and drawing out information using my phrasing, not the limitations of their index and table of contents.
-
As expressed in an earlier post it can produce a wide range of helpful output.
I’m sure I’m under-utilising it.
Other models
Occasional use via Duck.ai. It’s free, offers a wide choice of models, and they promise not to train on your data. Good for research and short-term efforts which you don’t want your main AI to ‘remember’ and distort its future efforts. For example, investigate something for a friend without etching that interest into your primary AI/LLM tool. Mind you it’s possible to run ChatGPT in ‘private mode’ and achieve the same result.
What I’m watching in 2026
A better Siri. Apple Intelligence is reportedly getting a boost from Google Gemini, which could finally make Siri useful beyond setting timers.
Open-source models like Qwen from China that can run locally on a laptop. Offline capability means privacy, no subscriptions, and something to tinker with on a long flight.