AuthorizedWork
Essays on how expert judgment becomes something a system — human or agent — is allowed to act on.
PRAXSO works with organizations whose consequential work waits for a few people, breaks between functions, or becomes unsafe when a capable AI system is treated as an authorized operator.
We capture the rules, exceptions, and decision rights those people hold, then build and operate the system humans and agents execute against — with explicit limits and a way back.
Start here: why capability is not authority. Then read in this order: why speed is no longer the constraint, who the work depends on, the record a system needs, how authority is granted, and why a customer promise has to survive the handoffs.
Start here
Capability Is Not Authority
Every company can now buy the same AI. The advantage goes to the companies that decide what it may do, prove what it did, and learn faster than everyone else.
By Zach Lendon · Co-founder & CTO, PRAXSO · October 2026
Reading order
When Routine Implementation Gets Faster, Understanding the Work Matters More
When producing an artifact takes less effort, the artifact can arrive before the organization has resolved the system around it.
PRAXSO Editorial
Published
Exceptional People Don't Scale. Their Expertise Can.
Your best people should not have to become permanent bottlenecks. Structure their judgment so humans and agents can execute it, verify outcomes, and stop at the right boundaries.
PRAXSO Editorial
Published
The New Interface Between Expertise and Software
The design challenge is to make the conditions of expert judgment explicit enough for a person to inspect and a system to use.
PRAXSO Editorial
Published
The Frontier Is Where You Experiment. It Should Not Be Where You Operate.
Explore new model capabilities at the frontier. Turn expert rules into a system humans and agents can operate through task-specific evidence, explicit permissions, reversible recovery, and an accountable promotion decision.
PRAXSO Editorial
Published
You Can't Transform an Organization One Department at a Time.
Local improvements matter, but an end-to-end customer commitment only changes when its definitions, context, ownership, and decision points survive the handoffs between functions.
PRAXSO Editorial
Published
Introducing attested: receipts, not screenshots
attested keeps the proof where your tests saved it, in your CI run, and gives the pull request a signed receipt instead.
Zach Lendon
Published
From the archive
Legacy essay · BlueFolders
I Stopped Borrowing My Identity From Companies
A personal reflection on how AI is reshaping professional identity from titles and logos to proof, creation, and earned credibility.
Zach Lendon
Originally published
Legacy essay · BlueFolders
The Proof Layer (Prelude): How AI-run companies earn trust without slowing down
Observability explains, receipts prove. Learn why AI operations need verifiable proof cards before trust breaks at scale.
Zach Lendon
Originally published
Legacy essay · BlueFolders
Spec Debt: The costliest new tech debt in the age of AI agents
Treat every line of human implementation as a spec alarm. Define rules (specs, policies, evals, risk tiers), let agents play the game, and erase spec debt.
Zach Lendon
Originally published
Legacy essay · AI Imagineer
Retrieve → Relate → Run: Why “Vectors + Relationships + Workflows” Is the Missing Pattern for AI Engineering
Vectors find. Relationships explain. Workflows deliver. Why a Context Mesh - vectors + relationships + workflows - is the missing pattern for AI engineering.
Zach Lendon
Originally published
Originally published
Originally published
From complexityto impact.
If a workflow waits for one person, a promise breaks between teams, or a capable system is about to be given authority nobody decided to give — bring us that problem.