← All postsBlog

Meet Anna - your daily AI news aggregator that runs entirely on your laptop

· Avery NXR

If you work in or around AI, you already know the daily problem: the field moves faster than any single newsletter can keep up with. You either subscribe to ten different newsletters and lose an hour every morning, or you accept that you're missing things.

Anna is one of the 7 production-ready agent templates that ship with Avery NXR. She's a daily AI news aggregator that scans the web every morning, classifies stories by topic and sentiment, and emails you a personalized digest at whatever time you want.

She runs entirely on your laptop.

What Anna actually does

Every morning at the time you specify, Anna:

  • Scans 15-30 sources you configure — TechCrunch, The Information, arXiv, Hacker News, X/Twitter accounts, your favorite Substacks
  • Pulls the new stories since her last run
  • Classifies each by topic (foundation models, infrastructure, tools, policy, enterprise adoption, research)
  • Tags sentiment (bullish, bearish, neutral)
  • Filters by your interests
  • Drafts a digest email organized the way you'd organize it
  • Sends it to your inbox before your first meeting

All of this happens on your machine. No prompt or article ever crosses to a third-party AI provider.

What's running under the hood

When you open Anna's graph in the Avery NXR agent builder, you see something like:

Scheduled trigger (8:00 AM)
  → Web scrape (15 sources)
  → Extract article content
  → Classify by topic (LLM call to local Qwen2.5-Coder 7B)
  → Tag sentiment (LLM call)
  → Filter by interests (rule node)
  → Compose digest (LLM call)
  → Send email (Gmail/SMTP/Resend)

Every node is inspectable. You can drop in your own logic, add a Slack notification when something hits a certain confidence threshold, or fork it to scan a totally different beat.

What it costs

Anna runs on your local model. The marginal cost per run is electricity — a few cents for the GPU time on your laptop.

A cloud-based equivalent that calls GPT-4 or Claude for each classification and the digest composition would cost roughly $0.30-$0.60 per daily run, or about $100-$200 per year per user. Anna costs the Avery NXR license and nothing else.

What this unlocks

Once Anna is running, two things become possible that weren't economical with cloud LLM workflows:

You can scale her up — run her hourly instead of daily, monitor 100 sources instead of 30, segment by team interests instead of just yours. The cost doesn't change.

You can fork her — make a "competitor news aggregator" Anna, a "regulatory news" Anna, a "fundraising news" Anna. Each runs in parallel. The cost is the same machine running more workloads.

Try Anna in under 5 minutes

If you've already got Avery NXR installed:

  1. Open the Agents tab
  2. Find Anna in the template grid (she's the productivity template tagged ANNA · PRODUCTIVITY)
  3. Click "Use this template"
  4. Plug in your email provider (Gmail, SendGrid, Resend, etc.)
  5. Configure your sources and topics
  6. Hit Run

She'll do a dry run on the spot. Tomorrow morning, you'll have your first real digest.

If you don't have Avery NXR yet, request access at avery.software. Free Desktop tier, no card required.

Why we built Anna first

The 7 templates ship pre-loaded because we wanted the first hour with Avery NXR to be productive, not exploratory. Anna is at the top of the grid because daily news aggregation is the most common "I'd build this myself if I had time" workflow we heard from early users.

Now you don't have to build it. Use Anna, fork her, or build something completely different — but you've got a working example to start from.