Tips and tricks
Practical, opinionated advice on getting good results from a desktop AI assistant. Most of it is about spending the right amount of computation on each task rather than reaching for the largest model every time.
Speed and cost
The two things people want to change first, and what actually moves them.
Get faster responses
The real contributors to latency, and which ones you control.
Reduce API costs
Where BYOK spending actually goes, and how to bring it down.
When to use high reasoning
A test for deciding when depth is worth the wait.
A model for each task
A three-profile setup that makes switching a single click.
Working well
Habits that make a desktop assistant genuinely useful rather than a novelty.
Better prompts
Prompting for short, frequent requests about work already in front of you.
Stay in your workflow
Configuration that makes asking a question cheaper than switching windows.
Organize conversations
Habits that keep a growing archive findable.
Local models for private work
A two-profile split that makes the private choice the easy one.
By profession
Concrete SideNote Pro workflows rather than general observations about AI and work.
For developers
Repository questions, error triage, and knowing when to escalate.
For product managers
Ambiguity hunting, proposal comparison, and documentation queries.
For consultants and analysts
Document comparison, decision matrices, and defensible output.
For marketers and writers
Rewriting in place, cheap variants, and reviewing your own drafts.