Back to the shelf

059

Sakshya – Machine Log Evidence Room

A local-first desktop app that watches machine log folders and presents a clean, auditor-ready evidence trail.

Brainstorming · Started August 2026 · System · Notion-First

That's the whole thing.

The site's sound is off. The narration still plays.

This reading didn't load. The transcript is below.

Read the transcript

Sakshya means evidence. It means witness.

I recorded this as a voice note in August twenty twenty-six. At the pharma company I work with, differential pressure sensors, washing lines, and autoclaves write data into PDFs and Excel files at intervals. Not every older machine has a full electronic audit trail. Auditors want the full record. Pulling it together is still manual. Data-integrity guidance treats that kind of hybrid system as high risk.

Sakshya would be a local-first desktop app. It watches the folders where those files land, indexes them, and gives a clean evidence trail: every record, every interval, intact. Browse by machine, date, or batch.

Starting with differential pressure, one drop folder. Then generalize. Nothing built yet. Follow along.

Narrated in my own cloned voice, generated with ElevenLabs v4.

Sakshya means evidence, or witness.

I recorded the idea as a voice note in August 2026. The problem I was sitting with: machines in a pharmaceutical facility write their monitoring data into PDF files or Excel sheets at set intervals. Differential pressure sensors, washing lines, autoclaves. Not every older machine carries a full electronic audit trail, access controls, or e-signatures. When auditors ask to see all the monitoring that has taken place, someone has to gather it by hand.

Data-integrity guidance treats that kind of hybrid arrangement, electronic machine output into paper or PDF, as high risk. A true copy has to preserve meaning and the full audit trail. Manual assembly is the opposite of that.

Sakshya would be a local-first desktop app that watches the folders where those files land and indexes them continuously. You would get a clean, auditor-ready evidence trail: every record, every interval, chain-verification status, a PDF preview per record, backup verification, browsing by machine, date, or batch.

I plan to start small: one machine type, one drop folder. Differential pressure. Then generalize.

Nothing is built yet. Status: Brainstorming. Follow along if this matters to you.

01 · The problem

Before it existed

What hurt

Machines write their monitoring data into PDF and Excel files, and some older ones keep no full audit trail. When an auditor asks for all of it, someone gathers it by hand.

Who uses it

  • QA team
  • Engineering
  • Auditors

02 · The flow

5 steps, in the order they happen

  1. 01 / 05

    Point at a drop folder

    You
    Chooses the folder where machine files land
    It answers
    Intended: watches it and indexes new files as they arrive
  2. 02 / 05

    Browse the evidence

    You
    Filters by machine, date or batch
    It answers
    Intended: lists every record and interval, in order
  3. 03 / 05

    Open one record

    You
    Selects a record
    It answers
    Intended: previews the original PDF beside its details
  4. 04 / 05

    Verify the chain

    You
    Checks a record’s status
    It answers
    Intended: shows chain-verification status per record
  5. 05 / 05

    Verify the backup

    You
    Compares with the backup copy
    It answers
    Intended: confirms the stored copy matches the original

03 · What was special

What it stands on

  1. Local-first

    A desktop app that reads folders already on site. No cloud step for the evidence.

  2. Start with one machine

    Differential pressure, one drop folder. Generalize only after that holds.

  3. Research came first

    The data-integrity grounding was filed before the idea had a name.

Show the auditor the data exactly as the machine left it.

Making-Of

The machine already wrote the evidence. Sakshya only has to present it intact.

A voice note, kept verbatim in Notion

  1. The world before

    Sensors, washing lines, and autoclaves write data into PDFs and Excel files. Not every older machine carries a full audit trail, and showing an auditor everything was manual.

  2. One day

    8 August 2026. A voice note: a local-first Windows app that watches the folder where those files land and presents them as an auditor-ready evidence trail.

  3. Because of that

    1. Data-integrity guidance was already researched and filed, which treats a hybrid machine-to-PDF arrangement as high risk.
    2. The scope was set small: one machine type, one drop folder, then generalize.
    3. The open shape was written down: chain-verification status, PDF preview per record and backup verification.
  4. Until finally

    Not yet. Sakshya is a brainstorm, and nothing is built.

  5. Since then

    Open: when to pick it up, and which machine to start with beyond differential pressure.

The evidence Process, hands, and artifacts

Process

  1. Origin

    Voice note: a local-first presentation layer for machine logs

  2. Grounding

    Data-integrity research already filed before the idea was named

  3. Scope

    One machine type and one drop folder first, then generalize

  4. Open shape

    Folder watching, chain status, PDF preview, backup check, auditor view

Hands

Akshansh

  • The problem, the local-first answer and the small first scope

Agents

  • Claude CodeLogged the idea and the regulatory research in Notion