ProductsGiga Context

Giga Memory

Shared memory for your whole company

Tools like ChatGPT and Claude remember things about you. Giga Memory remembers things about your company: what's been decided, who's doing what, what's still open, and how your team likes things done. Anyone on your team can ask about it, from Giga Web, Slack, ChatGPT or Claude, and only ever gets answers from what they're allowed to see.

In the Giga web app, your memory lives in Brain.

Personal AI memoryGiga Memory
Remembers one person's preferences and chatsRemembers your whole team's decisions, commitments and know-how
Lives inside one appSaved from and used in Giga Web, Slack, ChatGPT, Claude, and more
Answers "what do I like?"Answers "what did we decide, and who's on it?"
Private to one userShared through Tracks, or kept private until you file it

What Giga remembers

Every memory is one clear statement, sorted into one of seven types:

TypeWhat it isExample
DecisionSomething settled for the group, including an agreed plan"We're dropping the free tier"
CommitmentSomeone named has said they'll do something"Warda will finish September payouts by Oct 19"
IdeaA proposal nobody has agreed to yet"Faith suggests a bonus for comments over 20 likes"
RequestSomething asked for, or an open question"Should the inbox auto-file confident items?"
PreferenceWhat someone wants or thinks should happen, with how strongly they feel"Tony prefers short bullet summaries"
FactHow things are, including someone's view of how things are"New TikTok accounts can comment 5 times a day"
ProcedureHow to do something"To deploy: merge to main, then restart in Hatchbox"

Each memory keeps who said it, a short quote from the source where there is one, and a link back to where it came from.

Memory keeps itself current

Memory isn't a pile of notes. When something new comes in, Giga links it to what it already knows:

  • a new decision replaces the old one, and settles competing preferences
  • saying a commitment is done closes it
  • someone taking on a request answers it
  • a plan built on an idea marks the idea adopted
  • hearing the same thing again adds evidence rather than a duplicate
  • two people wanting different things marks the preference disputed

So when someone asks "what did we decide about pricing?", they get the current answer, with the history behind it.

How memory gets in

Giga only saves what will still matter later. Small talk, one-off checks and status that will be stale in a few days are left out.

For meetings, the transcript beats the summary. When you import a meeting, Giga takes memories and quotes from the transcript itself, and uses the meeting's AI summary only to work out who's who and who owns what.

Where memory goes: Tracks, or private until you file it

Memory belongs to a Track, and a Track's memory is only available to the people on that Track. That's how memory about the Acme deal never shows up for someone who isn't on it.

When you save something without choosing a Track, it stays private to you and waits in Brain → To file, with Giga's suggestion for where it belongs. For each item you can:

  • File it to the suggested Track
  • choose another Track
  • keep it private

Filing is sharing: once filed, everyone on that Track can see it. Giga suggests a Track but never files for you. For another person's view, it asks you to choose the Track yourself.

Inspect your memory

Memory is never a black box. In Brain you can:

  • browse current memory by type: Decisions, Commitments, Ideas, Requests, Preferences, Facts, Procedures
  • open any memory to see its quote, who said it, when, and every piece of evidence behind it
  • read the Evidence log, the full record of everything saved, in order
  • follow Brainstorms: a brainstorm saved from a chat, showing how ideas built on each other and where you ended up
  • search and filter everything from the sidebar

You can also correct memory by asking Giga: "Neil shipped that", "that's wrong", "park that idea". Giga records the correction first, so there's always a trail.

Ask your company anything

Because memory spans your whole team, Giga can answer questions no single chat history could:

  • "Where are we on the Acme renewal?"
  • "What did we decide about the free tier, and why?"
  • "What's Warda committed to this week?"
  • "What's still open on the website redesign?"
  • "How do we deploy?"

When you start work on a Track, Giga reads that Track's memory first, so it picks up where things left off.

Memory follows access

People only get answers drawn from memory they're allowed to see: memory filed to Tracks they're on, plus their own private memory. This works the same in Giga Web, Slack, ChatGPT and Claude.