Data Shield
Source material stays on the user's hardware while you explore.
How turbolapper - AFM works on your Mac.
Architecture, Apple Intelligence, benchmarks, demos, and the questions technical buyers actually ask.
turbolapper - AFM is Local Analytics: raw data files to a slide deck on your Mac.
This page covers how turbolapper - AFM (Apple Foundation Model) runs Local Analytics on a Mac: architecture, performance, Apple Intelligence as the one brain, and public demos. We measure cold starts - nothing kept warm between queries - so every answer pays full start-up cost. That is the number that matters on a laptop.
- Speed ~120ms plateau
- Brain Apple Intelligence
- Trust Data stays on your Mac
Method
How Local Analytics works in turbolapper - AFM
Heavy data work on hardware you control. Apple Intelligence is the brain. Answers show their work.
Local Analytics means the heavy data work happens on hardware you control. turbolapper - AFM is the analyst methodology hardwired into a native macOS app - ingest, prep, ask, and deliver a slide deck with provenance on your machine. The brain is Apple Intelligence (Apple Foundation Model).
Built for macOS 26 or later with Apple Intelligence on an Apple Intelligence capable Mac. Download / requirements
In practice, the workflow looks like this:
Data stays on the Mac
Your dataset stays on-device with the work.
Prep does the hard work
Clean names, documented views, hygiene, honesty flags - so answers are fast and inspectable.
Apple Intelligence reasons
On-device foundation model included with the product.
Verified output
Answers show their work with sources and plain confidence language. Deliver as a slide deck.
That design is why turbolapper - AFM stays private and flat-cost as you ask more: prep and analysis run where the data already lives.
Performance
Benchmark snapshot
Cold starts only - full start-up cost paid every answer. The number that matters on a laptop.
On a real county-scale dataset, from thousands of rows to a million, cold-start answer time for turbolapper rose early then plateaued around ~120ms at one million rows - while common in-process baselines climbed into seconds.
~120ms
Cold-start answer time at 1M rows on published demos.
Local answers stay interactive as data grows on published demos.
In-process baselines
Common table tools
turbolapper
Chart shows the shape of the result: turbolapper stays interactive as data grows. Numbers come from published demos and may be refined over time.
Architecture
Three Shields
How private-by-architecture shows up for security-minded evaluators.
Engine Shield
Analytics compute runs locally for interactive answers on prepared views.
Inference Shield
Reasoning runs via Apple Intelligence on the device. One included brain for turbolapper - AFM.
One brain
Apple Intelligence - the model is included
turbolapper - AFM (Apple Foundation Model) is the analyst methodology plus Apple’s on-device foundation model.
Analysis stays on the Mac with Apple Intelligence, so your data stays with the work and follow-ups stay flat-cost.
On-device foundation model
Apple Intelligence provides the brain on macOS 26+ with Apple Intelligence ready. Built for private, interactive Local Analytics.
What you get with AFM
Prep, ask, and deliver a deck with one included model path. Flat monthly pricing for the app; Apple Intelligence does the reasoning on your hardware.
turbolapper does the heavy prep work locally and delivers inspectable answers and a deck. Read the commercial framing on Economics and the capacity ladder on Pricing.
Principles
How turbolapper is designed
Three principles that shape the Mac app - privacy, prep, and answers you can check.
Private by architecture
Your dataset and its heavy processing stay on the machine. The app keeps data on your Mac - built for sensitive work on your own hardware.
Prep heavy, run light
Absorb complexity at prep - clean names, documented views, hygiene, honesty flags - so on-device reasoning can answer from prepared views on every ask.
Verified, not guessed
Answers are inspectable and reproducible. Confidence in words - in plain language. When the data is thin, confidence stays honest.
Verified
Estimated
Directional
Proof
Demos and articles
Read the write-ups here. Session videos and thumbnails live only on Community.
US markets & segments demo
Tile maps vs night lights vs Nielsen DMAs - and DIY personas from public data on a Mac.
Read write-up →Federal data demo
Method demos on open federal series - multi-series report in hours on your Mac.
Read write-up →Retail prep demo
Raw vs prepared views on public fashion-retail data - coverage, speed, and honest coverage when the data is thin.
Read write-up →NYC taxi demo
Hundreds of millions of open rides on a Mac (M2 · 32 GB). Method, hygiene, and performance stance.
Read write-up →Unit economics
Flat cost of curiosity as a warehouse complement - a companion alongside pitch.
Read economics →Team Data Crunch sessions
Session archive, Slack, and how a public series works.
Open Community →FAQ
Local Analytics, Apple Intelligence, and how turbolapper - AFM works
What is turbolapper - AFM? Is it a local AI analytics app for Mac?
turbolapper - AFM (Apple Foundation Model) is a native macOS app that turns raw data files into a slide deck using Apple Intelligence on-device. You ingest and prep data on your machine, ask questions in plain language, and get answers with provenance. Requires macOS 26+ and Apple Intelligence. Join the macOS waitlist for download access.
What does “Local Analytics” mean, and how is turbolapper different from cloud BI?
Local Analytics means analysis runs on hardware you control - for turbolapper - AFM, your Mac. Cloud BI and cloud “AI analyst” tools usually assume data lives in a vendor path and that exploration spend rises with queries, seats, or tokens. Keep the warehouse for centralized reporting; use turbolapper for private exploration on device. Economics
How is turbolapper different from traditional BI dashboards?
Dashboards show what already happened after someone built the report - and usually assume a specialist queue and a central stack. turbolapper - AFM is for the private session: bring raw data files onto your Mac, prep, ask in plain language, and leave with a slide deck and answers you can inspect. Follow-ups stay on a flat monthly plan. Keep the warehouse and dashboards for governed reporting; use turbolapper when you still need another cut for the room. Economics
Where does my data live when I use turbolapper - AFM?
On your Mac. Analysis and Apple Intelligence run on-device, so your dataset stays with the work. Privacy Policy
What model powers turbolapper - AFM?
Apple Intelligence is included as the brain for turbolapper - AFM. Prep, ask, and deliver with the on-device model that ships with the product.
How do I chat with my data privately on a Mac?
Install turbolapper - AFM, bring your files onto the machine, prep views so questions are trustworthy, then ask in plain language. Public method demos: Community, federal data, retail prep, NYC taxi.
Do I need SQL to use turbolapper - AFM?
No. After prep, you ask in plain language on your Mac. Good prep still matters - clean names, joins, documented views - so answers stay trustworthy. You do not need to write SQL for every follow-up once views are ready. See prep heavy, run light and the retail prep demo.
Can turbolapper analyze large datasets on a laptop?
Yes. Public demos include hundreds of millions of NYC open taxi trips on a Mac (demo hardware: M2 MacBook, 32 GB RAM), multi-year public retail tables, and multi-series federal work. Cold-start answer times on this page stay near ~120ms at 1M rows in published demos. Federal data · NYC taxi · waitlist
Is turbolapper a Snowflake alternative or warehouse replacement?
Keep the cloud warehouse for heavy centralized scale. Use turbolapper - AFM for follow-ups and private exploration on your Mac at a flat monthly price. Economics
Why can local AI data analysis cost less than metered cloud AI BI?
turbolapper plans are flat monthly. They differ by capacity and prep depth. Follow-ups stay on your flat plan. Pricing
What does “prep heavy, run light” mean?
You invest once in clean names, joins, documented views, and honesty flags so on-device reasoning can answer fast from prepared views every time. The retail prep demo shows raw tables vs prepared views: coverage, speed, and honest gaps when coverage is thin.
How does turbolapper show confidence in answers?
Answers are meant to be inspectable and reproducible, with confidence labeled in words (Verified / Estimated / Directional) - in plain confidence language. When the data is thin, confidence stays honest. See the confidence chips and demos on this page.
How does turbolapper keep AI answers grounded and honest?
Reliability comes from method, not marketing absolutes. You invest in prep so questions hit prepared views; answers are meant to be inspectable with provenance; confidence is labeled Verified / Estimated / Directional; and when coverage is thin, honesty flags say so instead of faking certainty. Apple Intelligence runs on-device over the work on your Mac. Deep dive: retail prep · confidence on this page.
What are the system requirements?
macOS 26 or later, a Mac that supports Apple Intelligence, and Apple Intelligence turned on. A native Mac experience. App / waitlist
How does private-by-architecture work for security reviews?
For a CTO or security reviewer: privacy is built into the design: Data Shield (source material on your hardware), Engine Shield (analytics compute local), Inference Shield (Apple Intelligence on-device; app keeps data on your Mac). Site and waitlist details: Privacy Policy. Enterprise evaluation: Contact.
Why does Local Analytics on Mac matter for security?
The analysis loop is built so your files, compute, and Apple Intelligence stay on the Mac you control - not an “upload the spreadsheet to a chat product” path. That shrinks the third-party surface for sensitive exploration. Full shields language: private-by-architecture · Privacy Policy.
Who is turbolapper - AFM for?
Finance and risk leaders who own cloud spend and care that proprietary data is not uploaded; analysts who need answers in the same session; researchers, consultants, and teams with sensitive datasets. More on the homepage; multi-seat paths via Contact.
What kinds of work is turbolapper best for?
Exploratory analysis on files you already have, board or leadership deck prep, private datasets that should not leave the machine, and stacking follow-ups without a question meter. Public method demos show the shape of the work: federal data, NYC taxi, retail prep, US markets. It is a Mac Local Analytics app (files in, deck and answers out) - not a live ERP command center or a replacement for your warehouse. Requirements
Where can I see demos, benchmarks, and Team Data Crunch videos?
Benchmarks and architecture are on this page. Write-ups: US markets & segments, federal data, retail prep, and NYC taxi. Live sessions and the full video archive: Community.
How do I get started?
Confirm requirements, then join the macOS waitlist, review flat monthly pricing, and watch or read public proof on Community and this page. For multi-seat or procurement, Contact the team.
turbolapper - AFM ships as a native Mac app
Requires macOS 26 + Apple Intelligence. Join the waitlist for download. Flat monthly pricing from free Starter through Enterprise. Or talk with us on Contact.