Hey folks! 👋 Just wrapped up my first month with MeetGeek and wanted to share my "where to start" takeaways, since I dove in headfirst.
For cost/ops-minded people like us, I'd start here:
* **Connect your main calendar** (Google/Outlook) and let it auto-record for a week. The default settings are decent.
* **Immediately check the AI summary accuracy** on a few technical meetings (mine were about K8s scaling and RI planning). Tweak the "detail" slider if it's missing key numbers or decisions.
* **Export a sample transcript** to see the raw data structure. It's clean JSON, easy to feed into other tools if needed.
Pro tip: Set up a dedicated folder in your drive for the exports and automate it early. I use a simple Terraform snippet to sync my exports to an S3 bucket for archive (cheaper than keeping them all in MeetGeek long-term).
```hcl
# Example: Using aws_s3_bucket with lifecycle rule for transcripts
resource "aws_s3_bucket" "meetgeek_transcripts" {
bucket = "meetgeek-transcripts-${var.env}"
lifecycle_rule {
id = "archive"
enabled = true
transition {
days = 30
storage_class = "GLACIER"
}
}
}
```
The big win is automatically tagging action items and saving me ~5 hours a month of note-taking. Let me know what you find useful!
#savings
Love that you jumped straight to automating the export! That's the exact mindset for ops folks. I've been using MeetGeek for sprint retros and incident post-mortems.
The JSON transcript is indeed clean, but one caveat: watch out for PII in ad-hoc internal meetings. I had to add a simple `jq` filter to redact specific keywords before the Terraform sync to S3. Something like:
```bash
jq 'walk(if type == "string" and contains("prod-db-password") then "REDACTED" else . end)' raw_transcript.json > safe_transcript.json
```
Also, tweaking that "detail" slider is crucial for technical talks. Found it was summarizing away important error codes and pod names until I cranked it up. The AI thought they were just random strings 😅
#k8s
Your Terraform snippet is a solid foundation for cost control, but I'd recommend adding an explicit IAM policy to that bucket resource to lock down access. The transcripts are often more sensitive than people assume.
If you're already comfortable with Terraform, the next logical step is to integrate the export trigger itself. MeetGeek's webhooks can push to a Lambda function, which can then apply your redaction filters and initiate the S3 upload, eliminating the manual export step entirely. This creates a true closed-loop system.
One caveat I've found: the default JSON structure is clean, but it doesn't include speaker identification metadata unless you've explicitly enabled and trained that feature in MeetGeek's settings. For post-mortems, knowing *who* said what is often as critical as the content, so that's a setting worth verifying early.
IntegrationWizard