Skip to main content

ZeroWidth

ZeroWidth cuts through the noise to find where AI changes your business, builds the software to ship it, measures whether it works, and keeps what you learn, so nothing is learned twice.

Start a conversation See how we work
  • Operators ZeroWidth has worked with:
  • Material
  • UC Berkeley
  • MASA
  • ID
  • Conic
  • Provana
  • Red Door
  • Kipu

Four ways we work together.

Each engagement starts with the systems your team already uses and the outcomes your business already measures. We bring the AI fluency, the pattern library, and the discipline to ship.

  1. Learn
    AI Bootcamp & Literacy Primers

    Get your team fluent in what AI can and can't do at work.

    Learn more
  2. Scan
    Opportunity Scan

    Figure out which AI ideas are worth building, and in what order.

    Learn more
  3. Pilot
    Rapid Pilot Foundry

    Build and ship a working AI workflow in six weeks.

    Learn more
  4. Scale
    Workflow Redesign & Scale-Up

    Redesign the workflow, prove the gain, scale what holds up.

    Learn more

Distributed Cognition

Writing on AI work: essays and field notes from the studio, what we're learning while we build, in long form.

ZeroWidth Software

Compass finds the work, Prism opens the options, Workbench builds it, Caliper proves it, and Ledger remembers whether you were right, which is exactly what aims the next pass.

The operator: zv1 works every instrument in this loop, and asks before it acts.

Learn more

Finds where AI belongs in the work, and ranks what to build first. Every pass starts here.

Learn more

Explores the possibility space while it's still cheap, then puts the real questions to real people before anything's a plan.

Learn more

Builds the thing: production flows your team can read, step through, and ship.

Learn more

Proves it works, with a hard number you can put in front of anyone.

Learn more

Remembers what you decided, what you expected, and whether you were right.

Learn more

The shared margin: where any of it gets sketched, with anyone, AI included.

Learn more

zv1

Tell it what you need. It works the tools; you approve every change.

zv1ZeroWidth
  • Turns your documents into a living map of how your business works
  • Builds working agents: prompts, tools, and the knowledge bases behind them
  • Designs test datasets and rubrics, runs the evals, reports the scores with receipts
  • Finds the person who knows what's missing and drafts the interview invite
  • Waits for your approval on every change: you see the exact edit on a card before anything lands
  • Knows which tool you're in, remembers past conversations, works from your phone

Compass

Find where AI belongs in your business.

CompassZeroWidth
  • Workflow + tacit-knowledge mapping
  • AI-led stakeholder interviews: send a link, the interviewer asks, the transcript becomes draft pages
  • Data + system readiness audit
  • Opportunity ranking by impact and feasibility
  • Sequenced implementation roadmap

Prism

Explore an idea, then find out what people really think.

PrismZeroWidth
  • Infinite canvas: branch an idea along any axis, cluster what emerges, ground it in cited research
  • Studies with the full kit: scales and NPS, choice and ranking, MaxDiff, matrices, screeners, demographics, display logic
  • AI-led interviews with adaptive follow-ups: qualitative depth without scheduling calls
  • Your own audience or a recruited panel, with quotas and weighting for a sample you can defend
  • Analysis you can trust: themes with the verbatims behind them, discovered segments, significance-tested cuts, printable reports
  • Recurring trackers that post readings to Ledger: each wave builds on the last

Workbench

Build the behavior: software your team can read and change.

WorkbenchZeroWidth
  • Composable prompts, tools, and agents
  • Knowledge bases your agents search at runtime: documents, spreadsheets, or an extracted knowledge graph
  • Step-through debugging across multi-step flows
  • Model-agnostic: swap providers without rewriting
  • Versioned, deployable, observable in production

Caliper

Turn “does it work?” into a hard number you can ship against.

CaliperZeroWidth
  • Datasets of items in the shapes your team works with: raw text, chat transcripts, Q&A pairs, file attachments
  • Reusable rubrics: name the criteria, set the scale, write the descriptions raters need to score consistently
  • Evals launched from a dataset and snapshotted against a rubric, so edits later don't change past results
  • Invite teammates by email or share a link with anyone: guests rate without joining the workspace
  • LLM-judge evals that re-score every change against the same rubric: human judgment sets the bar, the judge holds it over time
  • Results dashboard with per-item breakdown, rater agreement, and lowest-scored highlights

Ledger

Remember what worked, what didn't, and why.

LedgerZeroWidth
  • Decisions with pre-registered expectations: written before the evidence, settled against it
  • Lessons and observations captured in seconds, straight into the shared record
  • The beliefs page: what the workspace currently knows, each belief citing its entries
  • zv1 as scribe: consult before acting, record as decisions happen, settle when evidence lands

Napkin

Sketch it out with your team, and an AI that sees the picture.

NapkinZeroWidth
  • Pressure-aware freehand ink that feels like drawing with a real pen
  • Realtime collaboration with live cursors: share a link, anonymous guests draw instantly
  • Paste screenshots and images; move, resize, rotate, annotate
  • Reference cards for real workspace entities: flows, pages, evals, and decisions, with connectors that re-route as you rearrange
  • zv1 sees the sketch: ask it about the diagram you just drew
  • Export any sketch as PNG: the whole board, the current view, or a selection