ARTHUR - BUILD IN LIVE #3
Systems Expanding: Auth, Payments, Uploads, and the First Microservices
Every week, I’m opening up the build of Arthur — what we’ve shipped, what we’ve fixed, and what we’ve learnt.
No theatre. No hype. Just the real work of turning Arthur into a mentor, co-founder and team founders can rely on.
Edition #3 is the week where Arthur’s edges start to connect — identity, payments, uploads, and the early architecture for reusable microservices.
We’re still integrating the Jiva SDK, which takes a little time, but the rest of the system is moving forward fast.
If you’d like to get involved, sign up for early access or join the community.
Arthur’s world is getting bigger
This week was about stitching the foundations together so real founders can actually use the system — real accounts, real uploads, real usage.
Arthur is now moving from “smart chat” into a proper product surface:
identity
billing
document flows
uploads
and soon: multi-agent research + structured outputs
Here’s what we shipped.
✅ What we shipped
🔐 Authentication is now live
The identity layer is wired in through Cognito.
What works today:
Email + password
Google Sign-In
LinkedIn Sign-In
Persistent sessions
There are a few UX edges still to polish, but the foundation is correct: real accounts, real users, real identity.
Optional later: Microsoft + GitHub if we see demand from builders.
This unlocks personalised workspaces, saved documents, usage tracking and Fuel Unit billing.
💳 Stripe integration — subscriptions + Fuel Units
Stripe is now wired into the app in development mode.
This unlocks:
subscription tiers
Fuel Unit top-ups
card payments
usage tracking
The final layer — inference-based usage — will come as we integrate into Jiva’s SDK, so Arthur can meter real model calls and price them accurately.
Jiva isn’t the starting gun; it’s the next accelerator on top of a surface that already works.
📥 Uploads: pitch decks, websites & video
The upload pipeline is now in motion — the first part of Arthur’s input layer.
Arthur supports:
pitch deck uploads
website crawling + analysis
video pitch uploads or in-browser recording
Simplified flow:
Upload → S3 → processed by the right engine → structured feedback + next steps.
🎥 Video analysis — far beyond transcription
Arthur won’t just read your pitch — it will understand how you deliver it.
Each video is analysed across three signals:
1️⃣ Speech
Transcript + timestamps + clarity → used for story and structure feedback.
2️⃣ Delivery
Pace, pauses, filler words, tone → capturing confidence and energy.
3️⃣ Visuals
Framing, eye contact, lighting, slide clarity, movement.
All three are fused into a structured coaching-style report — as close as we can get to a real pitch coach sitting alongside you.
This analysis pipeline is being built as a standalone system so it can power:
FUSED
accelerators and incubators
government programmes
external pitch competitions
🏗️ User Account Panel — coming next
The founder dashboard is now scoped and will include:
subscription & billing
Fuel Unit balance + usage
document library
integrations (Google Workspace, LinkedIn…)
preferred login provider
This becomes each founder’s main control centre inside Arthur.
📄 Reports, plans & documents
Arthur’s document layer has two types — by design.
1️⃣ Interactive (conversation-driven)
The living parts of a startup:
GTM
sprints
ICPs
messaging
fundraising prep
hiring strategy
These need back-and-forth, context and iteration — not one-shot templates.
They rely on memory, tagging and conversation history.
2️⃣ Secondary (template-driven)
Once information is confirmed, Arthur can auto-generate:
investor briefs
governance docs
onboarding packs
policy docs
technical notes
research summaries
These are delivered first through Google Docs so founders can edit, collaborate and refine.
🔗 Your evolving data room
Every document Arthur creates feeds into one thing:
A live, always-updated, machine-readable data room.
Not a static folder.
Not a quarterly rewrite.
A continuously refreshed backbone of:
plans
metrics
decisions
narrative
investor-ready artefacts
Updated automatically as Arthur learns more about you and your company.
⏳ What we’re working on
🧩 Integrating into Jiva’s agentic stack
Arthur’s surface — auth, payments, uploads, docs, memory — is already in motion.
Integrating into Jiva’s SDK is the next accelerant.
From Jiva, we plug in:
the research API
consistent response schemas
stable routing between Jiva’s agentic stack → model layer → Arthur’s middleware
This unlocks:
deep, multi-agent research
structured plans & briefs
inference-based costing + Fuel Unit accounting
department-level document generation at scale
It’s the layer that switches Arthur’s internal organisation from “defined” to “fully active”.
🛠️ Modular system design
Under the hood we’re building reusable engines — not one-off features:
pitch deck engine
website engine
video engine
research wrapper
structured output formatter
These power Arthur for founders, and also:
FUSED global events
accelerators and incubators
government programmes
enterprise / partner integrations
Arthur is not a silo — it’s modular from day one.
🚧 What’s next
Stripe → live mode
Full upload pipeline (end-to-end)
Founder account dashboard in the UI
Templates for secondary documents
Mapping agent → workflow → output chains for the first GTM / finance / legal journeys
Arthur is becoming a real StartupOS
Arthur is an organisation that runs behind a conversation — one that thinks with you and builds with you.
This week wasn’t glamorous, but it laid the rails: identity, payments, uploads, accounts, and the first microservices.
Once Jiva lands, we stop sketching and start stress-testing the real machine.
FOR THE ❤️ OF STARTUPS
Help steer what we build next
I’m not building Arthur in a vacuum.
If there’s a workflow, bottleneck or ugly bit of founder work you want Arthur to take off your plate, tell me.
What you’re trying to do
What slows you down
What “done” would look like if Arthur nailed it
These go straight into the build plan for the next sprints.




The three-layer video analysis aproach (speech, delivery, visuals) is really clever. Most tools just transcribe, but capturing framing and pace feels like it could actualy replicate what a pitch coach notices. Curious how the Jiva SDK integration will change the respons quality once its fully wired in?