BlogNotes from building Savvy
· 5 min read
The best AI note taker depends on one question: bot or no bot
An honest comparison of AI note takers and meeting assistants in 2026, grouped by architecture rather than feature count: bot or bot-free, post-call or in-call, vendor model or your own.
comparisonmeeting-assistantprivacy
· 5 min read
Claude Code and Codex CLI use cases that are not coding
A coding CLI signed in on your machine is a general-purpose local model runtime with a shell interface. What that unlocks outside the editor, and what we learned using one as a meeting assistant's engine.
Claude CodeCodex CLIbring your own modelarchitecture
· 4 min read
Deepgram or AssemblyAI: picking a transcription provider for live meetings
Savvy needs one streaming transcription key and supports both providers. How the two compare on latency, streaming price, model choice and language handling for meeting audio.
transcriptionsetupproviders
· 4 min read
Granola alternatives for Mac: what to use when notes are not the problem
A look at alternatives to the Granola meeting app, grouped by what you actually want instead: a team archive, Windows support, live in-call guidance, or an app you can read the source of.
comparisonmeeting-assistantmac
· 4 min read
Otter.ai alternatives: pick by what you are replacing
Otter does three separable jobs, live captions, a searchable archive, and post-call summaries. Choosing an alternative is easier once you know which of them you actually use.
comparisonmeeting-assistantprivacy
· 4 min read
What "real-time meeting assistant" actually means
Four different products are sold as real-time meeting assistants: live transcripts, live summaries, generic conversation coaching, and guidance grounded in your own documents. They are not substitutes.
meeting-assistantreal-timecomparison
· 5 min read
Local-first meeting assistants vs bots that join the call
How a local-first, bring-your-own-model assistant like Savvy differs from a meeting bot: who is in the call, where audio goes, who pays, and what you give up.
comparisonlocal-firstprivacy
· 5 min read
Bring your own model, with the CLI you already pay for
Savvy has no model of its own; it shells out to Claude Code or the Codex CLI on your Mac. What that means for cost, terms, rate limits and what leaves your Mac.
bring your own modelClaude CodeCodex CLIcost
· 6 min read
How Savvy decides when to speak
How triggers, background scans, skip rules and grounding scores decide whether Savvy shows a card during a call.
how it workstriggersgrounding
· 5 min read
What to prepare before a negotiation call
A one-page brief for a negotiation call: objective, both positions, priorities, questions, concessions, red lines, prohibited claims, and how Savvy reads it.
preparationnegotiationbrief
· 5 min read
Recording a meeting: consent, etiquette, and why Savvy has no hidden mode
One-party and all-party consent for recording calls, what to tell the other side before you transcribe, and why Savvy has no hidden mode.
consentprivacyetiquette