The case for a deterministic deck
A serious deck shouldn’t be a gamble. Making one shouldn’t need a designer.
Every other tool makes you choose — fast or trustworthy, easy or on brand, beautiful or yours to keep. Lattice is the argument that you shouldn’t have to. A deck is plain text; a design system does the design; and the render you approve is the render that opens in the room.
This page makes that case in full: the problem, what we believe a deck should be, what we commit to — and an honest, sourced read on every alternative, including where Lattice is the wrong call.
The problem
Three decades of tools copied PowerPoint’s blank canvas — and its blank check.
A slide tool hands you an empty rectangle and no guardrails. Nothing stops the drift, enforces the brand, or catches the contrast that fails and the chart that’s unreadable in gray. The burden of consistency falls on you, slide after slide, the night before the meeting — and you can’t even tell what changed between two versions. So we all sit through decks that are worse than they needed to be, not because the author had nothing to say, but because the tool fought them the whole way.
AI didn’t fix the foundation. It automated the drift.
Bolting a model onto the blank canvas made the first draft faster and the finish worse. The output has a name now — “slop,” Merriam-Webster’s word of 2025 — the same gradient, the same font, the same rounded card, on every deck. Sameness is the tax you pay for design-by-generation.
The foundation was wrong for today’s work, and no amount of AI layered on top can mask a flaw that deep. The fix isn’t a smarter canvas. It’s no canvas at all.
Consistency is the tell
What separates a professional deck from an amateur one usually isn’t budget or talent — it’s consistency that holds from the first slide to the last. The moment it breaks, the reader stops trusting the argument. A design system keeps it for you, so the trust is never yours to lose.
What we believe
The system carries the consistency. You carry the judgment.
Most of what goes wrong in a deck is mechanical — contrast, spacing, alignment, a color that’s the only signal a color-blind reader can’t see. None of it is a matter of judgment, and none of it should cost you any. Lattice clears all of it off your desk automatically, so the only thing left in front of you is the thinking: whether the claim is true, whether the slide earns its place, whether the argument lands. The machine owns what’s correct; you own what’s good.
Good design, made a default
What reads as authoritative isn’t more design — it’s consistency and restraint, held on every slide. A design system makes that automatic: the polish that used to take a designer becomes the place you start from. You bring the judgment; the frame makes sure it looks the part.
A slide is four decisions
Function, Form, Substance, Finish — what the slide is for, the shape that holds it, what fills the shape, and the palette and mood. You make them one at a time instead of fighting all four at once on a blank canvas. Change one without disturbing the rest.
Built to be read without you
A board deck gets forwarded, attached, opened in a room you were never in. It has to carry its point alone — one clear claim, a heading that says something, a shape that fits the thought. Lattice is built for the deck that still makes the case after you’ve left.
On speed
Fast to a draft is not fast to a deck.
An AI generator is fast — to a first draft. That’s the familiar 80/20 of generative AI: the demo lands in a minute, and the last mile is still yours. Then the real clock starts, and it runs long.
Their clock
You fact-check the numbers the model invented. You strip the sameness so it doesn’t read as a template. You fix the export that came back with the wrong fonts. You force the whole thing back onto the brand. The draft arrived in a minute; the deck you can send is hours away — and the same hours come due on the next deck, and the one after that.
Ours
Lattice front-loads the cost once. You learn the tool and you write the words — and after that, every render is final: on brand, correct, identical each time. No re-brand, no broken PPTX, no fact-checking the layout, not on this deck and not on the next fifty. Their draft is fast and the fixing never ends; ours starts slower and the finish comes free.
The artsy deck wins the demo. The deterministic deck wins the boardroom.
What we commit to
Four promises the render path keeps.
Four guarantees the whole engine is built to hold. Everything else on this page follows from them.
Deterministic by contract
The same Markdown and the same theme render the same pixels — on your laptop, a colleague’s machine, a build server — this year, next year. In regulated work that’s not a nicety: reproducibility and auditability are what standards like ISO/IEC 42001 now ask for. It’s how you put your name on a deck without re-reading every slide: sign it off once, and that’s what ships — not a fresh render that drifted overnight.
The AI writes; the engine renders
A model can draft the words; it never touches the pixels. Probabilistic reasoning upstream, reproducible output down — the way regulated work keeps its reasoning layer and its record layer apart. AI generators fuse content, facts, and design into one roll of the dice and get each one only mostly right.
Your model, your account, your spend
Drive any of 500+ models through your own OpenRouter account, switch whenever you like, and watch the meter the whole time. The AI layer answers to you, not to us. The tool that generates your deck should not also be the one that bills you for it and swaps the model out whenever it likes.
Your work is yours to keep
Lattice is open source under AGPL-3.0, and a Lattice Output Exception frees every deck you make — no license attaches to your slides, ever. It runs offline, phones home to nobody, and can’t be shut down on you. Leave whenever you like: every deck is still a plain PDF sitting next to the Markdown that built it.
Now, the field
A fair read on every alternative — including where each one wins.
That’s the argument. Here’s the map. We credit each family’s real strengths and name where Lattice is the wrong call, because a comparison you can’t trust is worth nothing. The presentation world splits into four camps.
AI generators
Gamma · Beautiful.ai · Decktopus · Presentations.ai · Plus AI · (Tome, RIP)
Where it wins. They are fast to a first draft, and the category is real money — Gamma alone reached a $2.1B valuation and roughly 70 million users. A prompt becomes a presentable draft in about a minute, and it looks modern out of the box. Nothing kills a blank page quicker. For a throwaway internal deck, grab one.
Where it falls short. The draft is the easy part. Regenerate one slide and the layout reshuffles. Exports break: Gamma flattens its PowerPoint to uneditable images and swaps your fonts for Calibri. Every deck drifts toward the same house style, and the numbers arrive with no sources.
Office suites + AI
PowerPoint + Copilot · Google Slides + Gemini · Apple Keynote
Where it wins. These own the boardroom today. PowerPoint is the universal format, Google runs the best live co-editing on the market, and everyone already knows how to drive all three. That reach is a real moat, and it isn’t going anywhere.
Where it falls short. The file is a binary blob or a cloud object, so you can’t reliably compare two versions — you get final_v3_FINAL.pptx instead. Fonts fall back and layouts reflow on the next machine, so “looks right here” means nothing. The brand guide sits in a folder nobody enforces. And Copilot guesses like the rest — even bundled into Office, it’s paid for by only about 3% of seats.
Code engines
Marp · reveal.js · Slidev · Beamer · Quarto · Spectacle
Where it wins. Our own family. Free, open source, plain text, versionable. If you live in a terminal, these are the lock-in-free choice, and Slidev is the sharpest of the lot.
Where it falls short. None of them ships a design system. You get a handful of themes and raw CSS, then you build every layout yourself — and boardroom polish takes real design work most people can’t spare. Several use a browser-print PDF path, so the output changes machine to machine. They’re for developers; we’re for anyone who has to make the deck — the Studio runs the whole loop, write to present, in a browser tab.
Design & collab tools
Pitch · Canva · Figma Slides
Where it wins. Strong on design, strong on collaboration. A marketing team that wants beautiful slides and live multiplayer editing gets both here. Figma’s design control and Canva’s template range are the real thing.
Where it falls short. The deck lives in someone else’s cloud as a proprietary object. You can’t rebuild it from text, and you steer the brand through a GUI kit instead of a system. Canva output reads as “a Canva template.” Figma Slides still ships PPTX export bugs. And a tool in someone else’s cloud follows its owner’s priorities, not yours — Pitch pulled back sharply in 2024.
Side by side
The capabilities that decide a high-stakes deck.
We score capability, not feature count. Lattice loses rows here. It trades live collaboration and point-and-click ease for reproducible output, enforced brand, and a file you own. That trade is the whole pitch — and it’s the right one exactly when the deck matters.
| Capability | Lattice | AI generators | Office + AI | Code engines | Design / collab |
|---|---|---|---|---|---|
| Deterministic, reproducible output Same input → identical pixels, every time. | ● | ○ | ◐ | ◐ | ○ |
| A deck you can track and roll back Compare two drafts, see what changed, undo a bad edit. | ● | ○ | ○ | ● | ○ |
| Brand enforced by the theme itself Tokens own the pixels; nobody can drift off-brand. | ● | ◐ | ◐ | ◐ | ◐ |
| Boardroom polish with no design work Looks finished before you touch it — down to the finish layer. | ● | ◐ | ◐ | ○ | ● |
| Deck review built in The tool itself checks the deck and names the fixes — free, before any AI runs. | ● | ◐ | ◐ | ○ | ◐ |
| Export fidelity across formats It looks the same on every machine you send it to. | ● | ○ | ◐ | ◐ | ◐ |
| Real-time co-editing Several people in one live document. | ○ | ● | ● | ○ | ● |
| WYSIWYG, no syntax to learn Drag, drop, done — nothing to memorize. | ○ | ● | ● | ○ | ● |
| Reliably agent / AI-authorable An LLM drives it to on-brand output, not a guess. | ● | ◐ | ◐ | ◐ | ◐ |
| No lock-in (your file, offline, open source) It runs where you run it and can’t be taken away. | ● | ○ | ○ | ● | ○ |
Swipe the table sideways for the other four columns →
● Strong ◐ Partial / conditional ○ No Columns are families; individual tools vary — see the notes above and the sources below.
The receipts
The case, with sources attached.
Strong claims need evidence. Every claim below links out and carries the caveat a skeptical reader should want — including where the source is thin.
Citations correct in a Big Four firm’s own AI report
KPMG’s 2025 report on agentic AI cited 45 sources. Five were correct. The rest were fabricated or mangled by AI — invented authors, wrong dates, papers that don’t exist. A firm that audits others couldn’t audit its own deck. KPMG pulled the report.
GPTZero’s primary analysis, corroborated by The Register and the OECD AI Incidents monitor. The sharpest evidence that a fabricated fact ships at the speed of a draft.
GPTZero investigation, 2026 ↗A government report that invented its sources
Deloitte produced a AU$440k report for the Australian government that cited non-existent papers and a fabricated quote attributed to a Federal Court judge. Deloitte refunded the contract’s final installment and reissued a version disclosing it had used GPT-4o.
Verified across Fortune, The Register, and CFO Dive. The AU$440k is the contract value; the refund was the final installment, not the whole fee.
Fortune / The Register, Oct 2025 ↗How much of a mainstream AI deck checks out
One team fed the same 10-slide prompt to the tools most people would reach for and verified every claim against primary sources. Gamma scored 20%. Beautiful.ai and Canva managed 17%. Canva pinned invented statistics to a real, named school. A fast draft is still a draft full of things to check.
One author, one prompt per tool — read it as directional, not a large study. Two lesser-known tools in the same test scored higher (44% and 43%); the mainstream names clustered at the bottom. The mechanism holds: an LLM predicts plausible text; it does not look facts up.
Fact-check of 6 AI presentation makers, Feb 2026 ↗Decks that vanished with their platform
Tome reached more than 20 million users, then shut its presentation product down in 2025 because almost nobody would pay for the output. Decks nobody exported in time were gone for good. A Markdown deck in your own folder cannot be switched off under you.
Under 2% of users paid, on under ~$4M revenue against $81M raised. Reach is not the same as value — or permanence.
Founder’s post + VentureBeat, 2025 ↗If the receipts land, the proof is one click away — write a slide and watch it render, no sign-up. Open the playground — or keep reading: our own weaknesses are further down.
The fair question
“Isn’t Lattice just Marp with a coat of paint?”
No. We did start from Marp’s one good idea — write the deck in Markdown, render it through headless Chromium to PDF — building on its engine early, then replacing it with our own at v1.0.0. Plain text buys a deck you can track and compare, true of every code engine; determinism is rarer — several print through the browser and drift from machine to machine. And Marp itself stays deliberately skinny: no component system, no design tokens, no layout primitives — a few generic themes and raw CSS, by design. Fair for a framework; it just isn’t a design system. Lattice brings the rest:
What Lattice is now
- More than fifty components across thirteen buckets. Pick the meaning, get the composition.
- A design-token system and 14 editorial palettes + 4 color-blind-safe. Layouts never name a color.
- Boardroom-grade defaults and WCAG-AA contrast in every palette, light and dark.
- An owned, pinned render engine. It renders plain HTML sections, not SVG-wrapped slides, so modern CSS works natively — container queries, counters, masks — and layout holds across browsers.
- Agent authoring. A machine catalog and a shared linter give an LLM rails, so it composes valid, on-brand decks.
- The Studio. The whole loop in a browser tab — write with live preview, get coached, apply fixes from chat, present — nothing to install.
- A finish layer. Premium surface treatments — marks, washes, backdrops — applied with one line and tuned visually in Fabricate, the Studio’s finish designer.
On AI
Bring your own model. We don’t own the AI layer.
The AI generators own the model: theirs, priced their way, replaced whenever it suits them. Lattice takes the opposite stance: the AI layer is yours. You pick the model today, and you keep that choice tomorrow.
Your model, your choice
Lattice talks to OpenRouter, so you can drive any of 500+ models — Claude, GPT, Gemini, DeepSeek, Llama, and the free tiers — and switch whenever you want.
We build and test Lattice on Claude, because we think it’s the best, and we default to Sonnet for the best work per dollar. Changing it is one dropdown.
The meter stays in your hand
We use AI too, and we know how fast the meter runs. So the Coach finds and fixes most issues with no model call at all; when the AI does write, Lattice caches the instructions and sends a compact catalog instead of a wall of docs.
Pick a model with its live price in front of you, set a session budget — Lattice warns at 80% and can stop at 100% — and your spend and balance stay on screen the whole time. No surprise invoice.
Today you connect your own OpenRouter account in one click — your key, your credits, your spend. Next: bring your own provider key, so you can point Lattice straight at Anthropic or OpenAI and cut out the middle layer entirely.
Where we’re weaker
When Lattice is the wrong tool.
A page you can trust owns its weak spots. If any of these is a hard requirement, use something else. We’d rather lose you here than disappoint you later.
No WYSIWYG canvas
You write Markdown and pick a component; you don’t drag a box across a canvas. The Studio narrows the gap — live preview, an apply-a-fix chat, a visual finish designer — but the source of truth stays structured text. If you think in pictures instead of structure, that grates.
A real learning curve
Markdown, a vocabulary of layouts, and an authoring contract with rules. The tooling teaches as you go — autocomplete, a live gallery, an inline linter — but your first deck still takes longer than typing a prompt into Gamma.
No real-time co-editing
Collaboration is asynchronous: you review a change and fold it in, rather than sharing five live cursors in one document. That’s how serious documents get edited everywhere else — but if your team lives in Google Slides, it’s a real change of habit.
A young ecosystem, a real install
The Studio and the playground run in a browser tab with nothing to install, but rendering on your own machine still means installing a toolchain, and the desktop app that removes that last step hasn’t shipped. The engine is at v1 — components and tokens are stable; the tooling around them is young. You’re early.
The PPTX is pictures, not text boxes
Our PowerPoint export is images of the slides — pixel-faithful, font-exact, identical to the PDF, but not text-editable in PowerPoint. We ding Gamma for shipping flat images, so we own it here too. The difference: ours regenerates from source in seconds and never drifts; theirs is the only copy you get.
Not for the stage
No builds, no transitions, no presenter animations. Lattice makes documents you email and archive and that read without you in the room. For a live keynote with stagecraft, use Keynote.
Where Lattice is going
The deterministic core is the floor, not the ceiling.
Everything above ships today at v1. Every roadmap item runs on the same engine, with the same guarantees, and adds reach without giving any of it up.
SlideWright desktop
A native app wraps the same engine, for people who want the determinism and skip the terminal entirely.
Deeper agent authoring
The machine catalog and the shared linter make Lattice an easy target for an LLM — so an agent drafts a deck that lands on brand the first time.
More, held until it’s ready
We’re building more on this engine, and we’d rather ship it than tease it. Whatever comes next clears the same bar as everything above — deterministic, yours, on brand.
See for yourself
Don’t take our word for it.
Open the playground
Write a slide, switch every palette, export a real PDF. In the browser, right now.
Render your first deck
Install the toolchain and produce a boardroom-grade PDF in a few minutes.
Read the method ↗
Every claim on this page, with its sources, its caveats, and the parts we could not fully verify.
Method: we surveyed the field in 2026 and lean on primary sources first — official docs and pricing pages, named studies, first-party postmortems — then recent reviews. These tools move fast and some claims are contested, pricing and AI features most of all. Where we could not verify a number, we kept it qualitative. Spot something out of date? Send a correction ↗