A Cryptographic Identity Can Replace API Keys
Combining a persistent Nostr identity with FIPS networking could let software arrange access across apps without sending users through accounts, domains and copied credentials.
Read the snack ↗The archive
Every idea,
given room to work.
Each snack is a self-contained argument developed from a long-form conversation.
Episode 72
The Internet We Deserve ↗
AI Agents Can Orchestrate Local Creative Tools
An AI agent can turn mature applications on one computer into an adaptive production pipeline by choosing the right tool for each task and iterating towards a defined standard.
Read the snack ↗
FIPS Addresses Connect Local Wingman Apps
A FIPS-based Wingman prototype made laptop-hosted apps reachable from other devices without first provisioning the usual public web infrastructure.
Read the snack ↗
FIPS Makes Personal Computers Internet Participants
FIPS gives computers direct, secure addresses so their files and services can be reached without first moving them into the cloud.
Read the snack ↗
How FIPS Finds Routes Between Computers
FIPS lets peers find a path to a computer or application, then carries IPv6 traffic across whatever underlying connection is available.
Read the snack ↗
Local AI Agents Put Existing Computers to Work
Bringing intelligence onto a user’s device lets agents work with local software, files and processing power instead of recreating the whole environment in the cloud.
Read the snack ↗
One Website Can Index Your FIPS Drives
A Nostr identity can turn files held across several FIPS-connected computers into one browsable index without moving them into a hosted repository.
Read the snack ↗Episode 71
The AI Industry is Wrong ↗
Automation leaves humans the hard decisions
As AI absorbs routine work, people inherit a smaller stream of unresolved decisions that demands sharper attention and clearer presentation.
Read the snack ↗
Breaking expert work into repeatable steps
Complex expertise becomes easier to execute consistently when it is divided into small decisions that people, models and tools can handle separately.
Read the snack ↗
How agents create information markets
Machine customers can turn reusable information into a market by comparing the price of an existing artifact with the cost of creating it again.
Read the snack ↗
Turn agent discoveries into repeatable workflows
Agents are useful for finding a path through unfamiliar work, but repeatable execution belongs in a bounded workflow.
Read the snack ↗
Turn agent workflows into cheaper software
Agents create lasting value when the processes they discover are captured as reusable software that runs reliably at a fraction of the original cost.
Read the snack ↗
Use agents only for uncertain work
Reliable AI workflows keep known operations in ordinary code and call an agent only when flexible judgement is genuinely useful.
Read the snack ↗
Why AI accuracy isn’t business reliability
A polished AI answer can create more risk than value when a business needs work it can safely rely on.
Read the snack ↗Episode 70
The Malleable Computer ↗
A shorter route into every idea
Short, timestamped notes can bring more of a podcast into view while asking less of both the writer and the audience.
Read the snack ↗
AI is a capital good
The value of widely available intelligence emerges when people apply it to a real business, idea or point of pain.
Read the snack ↗
Care has to enter at both ends
AI can help shape creative work without hollowing it out, provided the creator supplies the substance, judges the output and remains responsible for the finish.
Read the snack ↗
Give every recurring job a visual home
As software becomes cheaper to create, recurring agent work can move beyond a universal chat into personal interfaces shaped around the job.
Read the snack ↗
Private access without public endpoints
Signed device access could make locally hosted Autopilot WApps available through Wingman without turning every installation into a public internet service.
Read the snack ↗
When a task outgrows its brief
Keeping exploratory work in conversation creates room for each useful question to expand the ambition of what is being built.
Read the snack ↗
When agent work stops coming back quickly
As agent jobs stretch from minutes into hours, working with them becomes less like a rapid exchange and more like managing several unfinished tasks at once.
Read the snack ↗Episode 69
AI Agent Economies ↗
Put intelligence where judgement matters
A hybrid monitoring workflow keeps sensitive data local, reserves frontier models for anomalies and turns repeatable answers into dependable software.
Read the snack ↗
Sell the AI outcome, not the stack
Businesses can adopt AI more easily when one provider translates a desired outcome into a complete working service.
Read the snack ↗
The agent leaves but the server stays
Cheap machine payments create a dangerous asymmetry when an anonymous agent can use infrastructure whose operator remains identifiable and accountable.
Read the snack ↗
Turning experience into software
AI can give one person the capacity to package years of expertise or solve a small personal problem, then distribute the result far beyond its original setting.
Read the snack ↗
When an agent can pay
Giving an agent limited spending power lets it complete routine purchases without surrendering human control over consequential payments.
Read the snack ↗
When an agent rents its own server
A wallet gives an AI agent the means to rent infrastructure, move its runtime and pay to keep itself online.
Read the snack ↗
When research agents cross the line
An agent that gathers and analyses its own evidence can turn an ordinary research request into an unauthorised intrusion.
Read the snack ↗Episode 68
One-Person Startups ↗
The founder’s portfolio
Cheap AI experimentation lets a solo founder spread entrepreneurial risk across several business ideas instead of staking everything on one company.
Read the snack ↗
The replication advantage
AI lets small software entrants reproduce established products at far lower cost, creating a temporary opening to attack margins once protected by expensive technology.
Read the snack ↗
The slow drain on software margins
Cheap AI replication gives lean entrants an opening against high-margin software incumbents, but slow customer switching could stretch the contest over years.
Read the snack ↗
When a product listing costs more than the stock can bear
Live commerce makes awkward inventory economical to sell by moving it through a live stream.
Read the snack ↗
When coding agents became a team
In 18 months, coding agents took Rod from needing someone else to build his ideas to supervising parallel teams that review their own work.
Read the snack ↗
When small businesses build their own software
AI-built internal tools could let small businesses keep efficiency gains that once flowed to software vendors.
Read the snack ↗
When startup capital costs control
Rod’s experience shows how raising money to build a technology company can steadily turn a founder with control into an executive accountable to other owners.
Read the snack ↗Episode 67
Local AI Models ↗
Don’t make the agent run every step
AI automation becomes more efficient when deterministic operations stay in code and agents appear only where the work requires judgement.
Read the snack ↗
The bandwidth limit on local AI
A model may fit in a local AI machine yet still respond slowly if its memory cannot supply data quickly enough.
Read the snack ↗
When agents need a shared layer
Flight Deck brings delegated work into one communications layer where tasks remain visible, missed activity can be recovered and access follows clear boundaries.
Read the snack ↗
When bigger competitors help
For a small company in an unfamiliar category, larger rivals can bear the cost of teaching buyers before differentiation becomes decisive.
Read the snack ↗
When encrypted sync costs too much
Application-level encryption can impose more design complexity than it earns when an app already runs on its owner’s server.
Read the snack ↗
When your news builds a memory
Pete’s personal newspaper turns accumulated coverage into a connected resource that his AI agents can consult for context without burdening him with another unread queue.
Read the snack ↗
Why consumers reach for hosted AI
Local models can handle everyday personal tasks, but hardware costs and limited awareness leave hosted frontier AI as the easier consumer choice.
Read the snack ↗Episode 66
The Coldcard Hack ↗
A crisis desk that changes pace
An agent can filter a fast-moving crisis into technically verified updates, then ease its reporting cadence as the urgency recedes.
Read the snack ↗
Map the gaps in what you know
An agent can extend a personal knowledge map just far enough to reveal the next useful ideas, sources and connections.
Read the snack ↗
Pliny builds knowledge beyond the news feed
Andy designed Pliny to turn relevant updates into researched briefings that deepen as its record of his interests grows.
Read the snack ↗
The economics of an attack swarm
AI makes repeated cyberattacks cheaper, while giving defenders ways to detect their computing footprint and raise the price of access.
Read the snack ↗
What personalisation leaves out
A briefing shaped around known interests can deepen understanding while making unrelated but valuable ideas less likely to appear.
Read the snack ↗
When cyber safeguards block defenders
A model that refuses exploit work can also deny incident responders the tools they need to locate a known vulnerability and check other systems for the same flaw.
Read the snack ↗
When randomness became predictable
A Coldcard firmware flaw turned private keys from secrets protected by an immense search space into numbers that attackers could recreate within hours.
Read the snack ↗Episode 65
Working with AI Agents ↗
Keep agent work small enough to steer
Agent work is easier to supervise when each job has a visible boundary and the human can intervene before too many decisions accumulate.
Read the snack ↗
Let an agent read the feed for you
A personal agent can replace an open-ended trawl through social media with one timely story chosen from a connected view of the news.
Read the snack ↗
Put the workflow in code
A pipeline can contain agent judgement without surrendering control of the wider process to an open-ended agent loop.
Read the snack ↗
Software shaped around the business
A shared technical foundation makes it practical to replace generic tools with applications that encode how one company actually operates.
Read the snack ↗
When four sessions blur the work
Spreading attention across four agent sessions and three projects can turn building into a hard-to-follow cycle of prompts.
Read the snack ↗
When revenue still depends on you
A business can earn millions and still remain a time liability when its operation depends on a founder continually sending instructions.
Read the snack ↗
Who gets to command the agent?
Signed messages give agents a verifiable way to distinguish an authorised operator from everyone else in a shared chat.
Read the snack ↗Episode 64
Vibe Coding with Power Tools ↗
AI coding agents work like power tools
Coding agents shift a developer’s attention from writing every line towards architecture, testing and review.
Read the snack ↗
Coding models have working personalities
Every coding model brings its own mix of literalness, initiative, caution and recurring habits to a coding session.
Read the snack ↗
Domain experts can build software from operational knowledge
Coding agents let operators turn their knowledge of workflows, exceptions and commercial pressures into working software.
Read the snack ↗
Long agent runs depend on reliable feedback
Tests, specifications and human review determine how long a coding agent can keep producing useful software on its own.
Read the snack ↗
Real use reveals missing product requirements
Working software exposes missing permissions, awkward screens and workflow exceptions that an initial specification rarely captures.
Read the snack ↗
AI speed makes architectural judgement more important
Rapid implementation lets an early choice about data, queues or background workers spread across a codebase within weeks.
Read the snack ↗
An agent harness controls context, tools and feedback
A coding harness controls which files an agent can inspect, which commands it can run and which results return to the model.
Read the snack ↗Episode 63
AI Is Great for Small Business ↗
AI can’t scale an adviser’s attention
Faster processing can expand a service business, but growth still runs through the scarce human capacity to pay attention and make sound judgments.
Read the snack ↗
AI gives financial advisers more client time
Regional Prosperity uses AI to increase back-office throughput so its people can devote more attention to the relationships that make its financial advice valuable.
Read the snack ↗
AI must solve immediate SME bottlenecks
For a small business under delivery pressure, AI becomes useful when it clears a bottleneck quickly enough to create value today.
Read the snack ↗
Build SME software around existing workflows
AI-assisted software building gives SMEs a practical alternative to reshaping their work around generic systems that only partly fit.
Read the snack ↗
How SMEs test and fix business bottlenecks
Small firms gain speed by testing ideas cheaply and directing investment towards the constraint that is holding back the business now.
Read the snack ↗
Technology gaps complicate regional business succession
Regional businesses of considerable scale can sit at opposite ends of the technology spectrum, creating very different challenges when their systems pass to a new owner.
Read the snack ↗
Why SME succession matters to regional communities
When most of an owner’s wealth sits inside a regional business, an absent succession plan puts both a family’s future and a community asset at risk.
Read the snack ↗Episode 62
AI for Freedom Tech ↗
AI Automation Increased Work Instead of Free Time
AI tools accelerated each task, but the endless cycle of prompting, testing and new releases made it harder to stop working.
Read the snack ↗
AI Coding Tools Need Cryptographic Guardrails
AI can accelerate software prototypes, but cryptographic code exposes why expert developers must still control the design and review every decision.
Read the snack ↗
AI Compresses Product Iteration From Weeks to Minutes
Coding agents can accelerate a familiar cycle of design, delegation and testing, but the faster cadence makes active oversight more important.
Read the snack ↗
AI Makes MVP Experiments Faster and Cheaper
AI lets one person turn a product idea into a testable prototype before recruiting a team or committing substantial capital.
Read the snack ↗
AI Writes Examples, Not Core Library Code
Library author Lighthazard uses AI to expand the supporting material around his work while keeping the code he cares about under his own control.
Read the snack ↗
Smaller AI Models Work Better on Narrow Tasks
Dividing software work into bounded jobs can make cheaper models a practical alternative to a single request sent to the largest available model.
Read the snack ↗
Wingman Turns Natural Language Into Agent Workflows
Pete built Wingman to turn a plain-language description of a process into an editable pipeline of code, specialised agents and loops.
Read the snack ↗Episode 61
AI Loop Engineering ↗
AI Agent Loops Work Like Managed Teams
The useful work in agent automation is defining the goals, operating rules and handoffs that let specialised agents coordinate without constant instruction.
Read the snack ↗
Better AI Application Beats More Model Intelligence
When technical experimentation outruns the task, simpler systems can create more value from capabilities that already exist.
Read the snack ↗
Compression Cuts AI Agent Context Costs
Filtering repetitive tool output before it reaches an AI agent can cut token use while preserving the context that matters.
Read the snack ↗
Human Checkpoints Keep AI Agent Work Aligned
Fast agent workflows create more room for iteration, but human judgement still has to redirect them before small choices compound into drift.
Read the snack ↗
Judge AI Spending by Its Return
When AI turns a predictable software bill into variable consumption, businesses need to measure the extra value each increment of spending buys.
Read the snack ↗
Training AI Agents Reduces Key-Person Risk
Managing AI agents like new workers turns operational know-how into reusable business systems that improve their output and reduce dependence on individuals.
Read the snack ↗
Why More AI Output Isn't More Value
Agent loops can multiply activity at remarkable speed, but the work only matters when it produces something specific, useful and wanted.
Read the snack ↗Episode 60
Models or Harnesses ↗
AI Agents Research Twenty Blockchains in Parallel
A shared evidence search can fan out across many targets before separate agents pursue the most relevant lines in depth.
Read the snack ↗
Cheap Inference Squeezes Frontier AI Economics
Frontier labs must keep funding expensive model advances even as older capabilities become cheaper and harder to distinguish from the latest release.
Read the snack ↗
Code-Directed Workflows Keep AI Agents on Track
Declarative pipelines reduce AI agent drift by fixing the path through functions, steps and data while leaving people responsible for direction and course correction.
Read the snack ↗
How AI Doomerism Could Shape Regulation
Warnings that AI could wipe out jobs can make the technology feel like an urgent political crisis, giving leading labs more influence over the rules that follow.
Read the snack ↗
Mathematical AI Must Separate Discovery From Retrieval
A model can find an unexpected link between mathematical fields, but novelty depends on proving it hasn't merely recovered an overlooked result.
Read the snack ↗
Purpose-Built Robots Beat Humanoids at Home
The strongest case for household automation is a machine that reliably removes a defined chore, not a humanoid whose broad capabilities may never justify its cost.
Read the snack ↗
Why AI Agents Still Need Human Direction
An AI agent can check its progress only against a destination it understands, and defining that destination may require much of the original work.
Read the snack ↗Episode 53
Own Your AI Stack ↗
Design starts before the interface
Agents can compress weeks of product review by exposing gaps in a loose brief before anyone commits to a polished interface.
Read the snack ↗
Faster work isn’t system change
AI can accelerate each employee’s workload while leaving the business’s repeated processes and handoffs untouched.
Read the snack ↗
Put repeatable checks on rails
Reliable execution comes from making required checks part of the process, rather than instructions that a person or agent must remember to follow.
Read the snack ↗
Smarter systems beat smarter models
The next leap in business automation will come from turning capable models into reliable parts of an operating system.
Read the snack ↗
When more agents mean more management
An agent workforce can automate a great deal of activity while still losing its cost advantage to software that delivers the same outcome with less supervision.
Read the snack ↗
When software becomes the business
As agents absorb more operating work, a firm's ability to own and move its encoded processes becomes a condition of business control.
Read the snack ↗
Your secret sauce is a text file
A firm can turn accumulated judgement into an owned asset by encoding how it works in private instructions that agents can use.
Read the snack ↗No snacks match those filters.