Apps for the work behind the work
Our product lineup. Stages describe work in progress; the sketches
illustrate each concept.
> cat
apps/spec-dude.md
spec-dude
Exploration
make intent visible.
A living map of what software should do. Connect discovery
conversations to requirements, screens, workflows, APIs, and
data. Explore specification coverage, surface ambiguity, and
make gaps understandable to product teams.
Assistant-native
Available as a Claude plugin, a ChatGPT plugin, and a
standalone assistant-native web app.
Use a prompt to run workflows and modify or update the app as
your needs change.
Explore the approach at
postui.org.
concept sketch / spec-dude
from conversation to coverage
-
DiscoverConversations, examples, open
questions
-
SpecifyScreens, actions, data, permissions
-
VerifyIntended behavior connected to evidence
An illustration of the idea, not a live app.
> cat
apps/auth-dude.md
auth-dude
In development
who can do what — and why.
Authorization policies that make roles, resources, and
permissions easier to reason about. Inspect effective access,
understand which rule applies, and review policy changes before
they reach an application.
Assistant-native
Available as a Claude plugin, a ChatGPT plugin, and a
standalone assistant-native web app.
Use a prompt to run workflows and modify or update the app as
your needs change.
Explore the approach at
postui.org.
concept sketch / auth-dude
make access explainable
- WhoRoles and groups
- WhatResources and capabilities
- WhyThe policy behind each decision
An illustration of the idea, not a live app.
> cat
apps/training-dude.md
training-dude
Exploration
turn what changed into what to learn.
A training concept that turns approved documentation, team
knowledge, and software changes into role-specific lessons,
knowledge checks, and in-app walkthroughs. Help people
understand a release and use it confidently.
Assistant-native
Available as a Claude plugin, a ChatGPT plugin, and a
standalone assistant-native web app.
Use a prompt to run workflows and modify or update the app as
your needs change.
Explore the approach at
postui.org.
concept sketch / training-dude
knowledge that follows the release
-
SourceApproved knowledge and product changes
- LearnLessons tailored to a role
- ApplyWalkthroughs and knowledge checks
An illustration of the idea, not a live app.
> cat
apps/crm-dude.md
crm-dude
Prototype
tell it what happened. know what is next.
A contact-centered CRM for small businesses. Bring client
activity, tasks, calendars, and follow-ups together, with an
assistant designed to propose updates for your confirmation.
Assistant-native
Available as a Claude plugin, a ChatGPT plugin, and a
standalone assistant-native web app.
Use a prompt to run workflows and modify or update the app as
your needs change.
Explore the approach at
postui.org.
concept sketch / crm-dude
keep the client in view
-
CaptureA conversation, a commitment, a next
step
- ReviewThe assistant proposes an update
-
ConfirmYou decide what goes into the record
An illustration of the idea, not a live app.
> cat
apps/notedude.md
notedude
In development
give working knowledge a home.
Capture and organize notes with Markdown and tags. Keep ideas,
decisions, and reference material close at hand, with a focus on
quick capture and a direct editing experience.
Assistant-native
Available as a Claude plugin, a ChatGPT plugin, and a
standalone assistant-native web app.
Use a prompt to run workflows and modify or update the app as
your needs change.
Explore the approach at
postui.org.
concept sketch / notedude
a place for the working context
-
CaptureThe thought before it disappears
- OrganizeNotes and tags
- ReturnDecisions, references, and ideas
An illustration of the idea, not a live app.
> cat
apps/todude.md
todude
Prototype
today, short term, long term.
A focused task workspace that connects immediate work to what
comes next. Organize priorities, add the context a task needs,
and keep the everyday list easy to act on.
Assistant-native
Available as a Claude plugin, a ChatGPT plugin, and a
standalone assistant-native web app.
Use a prompt to run workflows and modify or update the app as
your needs change.
Explore the approach at
postui.org.
concept sketch / todude
the next thing, in context
- TodayWhat needs attention now
- Short termWhat is coming next
- Long termWhat you are working toward
An illustration of the idea, not a live app.
Client work
Systems app_sciences architects and builds for other companies.
> cat
client-work/vendorbuddy.md
vendorbuddy
In development
vendors, minus the spreadsheet.
A client and vendor system built the boring, reliable way. nbino
is CTO and architect — currently weighing Firebase Hosting,
Functions and Data Connect against Vercel and Supabase.
Assistant-native
Available as a Claude plugin, a ChatGPT plugin, and a
standalone assistant-native web app.
Use a prompt to run workflows and modify or update the app as
your needs change.
Explore the approach at
postui.org.
> cat
client-work/algotrade.md
algotrade
Planning
strategies as a language.
An algorithmic commodities trading platform built around a
domain-specific language for expressing strategies, with
translators from NinjaScript and friends. Discovery and
architecture for a non-technical owner.
Assistant-native
Available as a Claude plugin, a ChatGPT plugin, and a
standalone assistant-native web app.
Use a prompt to run workflows and modify or update the app as
your needs change.
Explore the approach at
postui.org.
▞▞
algotrade
on the drawing board
> cat
client-work/predictmarket.md
predictmarket
Planning
markets for anything.
A blockchain prediction market where users create their own
markets. Architecture across the streaming spine, the CDN front
end, and the cost of keeping all of it running.
Assistant-native
Available as a Claude plugin, a ChatGPT plugin, and a
standalone assistant-native web app.
Use a prompt to run workflows and modify or update the app as
your needs change.
Explore the approach at
postui.org.
▞▞
predictmarket
on the drawing board
SDLC Orchestration
AI across the software development lifecycle, guided by explicit
intent, connected evidence, and people who own the decisions.
One thread from discovery to delivery.
A prompt is a starting point. Our methodology connects the
conversation, the specification, the implementation, and the proof
that it works.
The measure is specification coverage: which intended behaviors
are defined, implemented, and verified — including the awkward
cases.
The thinking behind spec-dude
01Facilitate discovery
Start with the people doing the work. Turn discovery
transcripts, examples, constraints, and open questions into a
shared understanding. Keep assumptions visible and decisions
traceable.
Outcome: Agreed intent and unresolved questions.
02Build a living specification
Describe screens, actions, permissions, APIs, data changes,
edge cases, and workflow intersections. Map coverage so
product owners can see gaps and clarify missing behavior
before it becomes code.
Outcome: A reviewable spec with acceptance criteria.
03Write the proof first
Derive tests before implementation from the specification.
Check that they fail for the missing behavior, then use unit,
integration, and end-to-end checks where each adds evidence.
Outcome: Requirements connected to executable checks.
04Implement and review independently
Build bounded changes and run the checks. Use a different
model to challenge the implementation, with human engineering
review covering architecture, security, maintainability, and
the intended behavior.
Outcome: Reviewed changes and passing checks.
05Validate the actual release
Deploy the reviewed change to a testing environment. QA
validates the exact build against acceptance criteria and
regression checks, recording the result and evidence. Release
authorization remains a separate decision.
Outcome: An identified build with recorded QA results.
06Release, teach, and learn
Publish the approved build with release notes and guidance for
the people using it. Feed observed behavior, support issues,
and new discoveries back into the specification and the next
round of tests.
Outcome: A release people can use, and a spec that keeps
learning.
Engineering approval
Is the implementation sound and ready to test?
QA acceptance
Does this specific build satisfy the agreed behavior?
Release authorization
Is this accepted build approved to go live?