Latest··5 min read
Strategic Intent as Part of the Agent Environment
Persistent operator intent and peer delegation need tests that measure useful work, preserved authority, and the human effort still required.
Observations from the work
Field Notes.
What we are learning about AI and the work that runs your business.
- ·7 min read
What Grok, Claude, GPT, Gemini, and DeepSeek Actually Trained On
Every frontier model starts from the same public web. The real difference is the private pile each lab glued on top, and what that means for your data if you build on one.
ai-vendor-selection·data-privacy·model-training·ai-governance·grok·claude·gpt·gemini·deepseek - ·12 min read
GPT-6 Astra and the Verification-Program Pattern: What OpenAI's Launch Shares With Fable 5.1
OpenAI's GPT-6 Astra ships gated like Anthropic's Fable 5.1, behind a verification program. The one independent benchmark that measures both says Astra isn't smarter.
openai·gpt-6-astra·daybreak·arc-agi·ai-native-sdlc·model-governance - ·9 min read
Faster Models, Slower Bottlenecks: Reading Fable 5.1 Against the AI-Native SDLC
Anthropic's Fable 5.1 and Mythos 5.1 make the build stage cheaper and faster. The AI-native SDLC playbook explains why that just moves the real bottleneck.
anthropic·claude·fable-5.1·mythos-5.1·ai-native-sdlc·agentic-coding - ·4 min read
Unknown is not the same answer as no
A status field that can only say pass or fail will lie by omission the moment the real answer is 'not computed yet.' Treating pending as blocked trains people to ignore both.
ai-operations·loop-engineering·ai-verification - ·3 min read
More writers does not mean more throughput
Say you give twenty autonomous agents write access to the same shared state. The bottleneck does not move to the model. It moves to whoever has to reconcile the collisions.
multi-agent·ai-operations·loop-engineering - ·3 min read
The check that cannot fail is not a check
A safety gate that has never been triggered looks identical to one that is broken. The difference only shows up when you go looking for the trigger it should have caught.
loop-engineering·ai-verification·methodology - ·6 min read
What business process automation actually costs
Almost nobody publishes a price for automation work, and the reason is not evasion. Here is the cost structure underneath a quote, and how to read one.
workflow-automation·ai-operations·buying-guide - ·8 min read
Persistence Doesn't Buy Better Judgment. It Buys the Right to Say No.
Every published agent harness spawns a worker, gives it a task, and reaps it. Ena Pragma runs named, persistent agents instead, and the payoff isn't more context. It's standing to refuse.
ai-operations-integration - ·4 min read
READY is not LIVE: the stages a loop must name
Coverage, readiness alerts, and public proof are different jobs. A green scorecard that confuses them will celebrate the wrong win.
loop-engineering - ·4 min read
Operating an autonomous company is an ops problem, not a magic agent
Autonomy fails when teams add agents without state, gates, verifiers, and receipts. Here is the operating shape that keeps a company in control.
operating-an-autonomous-company - ·6 min read
A running terminal is not a receipt
We wired inbound messages to coding agents by typing into a terminal. It worked immediately, which was the problem. Here's the transport architecture that replaced it, and the three receipts that never imply each other.
architecture·agents·reliability - ·4 min read
What's the difference between a project build and ongoing support?
A build has a finish line. Support does not. How the two engagements differ in scope, risk, and what you own at the end, and how to tell which one you need.
software-maintenance·methodology - ·2 min read
The charge that almost happened twice
A payment webhook with no idempotency guard, one retry away from double-charging real customers. How it surfaced and what the fix actually required.
case-study·verification - ·3 min read
The migration that almost ate a taxonomy
A production migration was about to trust a line of documentation that was wrong. What we caught, what we did about it, and why the catch is the product.
case-study·verification - ·2 min read
How do you take over someone else's code?
Inheriting a codebase nobody documented. How a takeover actually works, what gets assessed first, and why a rewrite is usually the wrong instinct.
technical-rescue·software-maintenance - ·2 min read
How Ena Pragma onboards a new maintenance client
Four phases: diagnostic, architecture, build and deploy, then ongoing operations. What happens in each, and why the order matters.
methodology·software-maintenance - ·5 min read
Software maintenance consulting
Hiring an outside team to keep software you already own running, rather than to build something new. What the engagement covers, and what it is not.
software-maintenance·ai-operations-integration - ·2 min read
What happens when your developer leaves?
The person who understood your system is gone. What actually breaks, what it costs to recover, and how to take ownership back without a rewrite.
software-maintenance·technical-rescue - ·3 min read
Notion Is Not a Note-Taking App
The folder habit dies hard in Notion, and it's the wrong mental model. The real shift is asking 'what kind of thing is this' instead of 'where does this go': property vs. location, not documents vs. folders.
knowledge-management·methodology - ·5 min read
Loop engineering is a reliability problem, not an intelligence one
Agent loops are not failing because the models are weak. They are failing because nobody built the state, verifiers, gates, and receipts around them.
loop-engineering - ·4 min read
Installing an agent skill is running untrusted code
A skill looks like documentation, so teams install one like they trust a README. The audits say treat it like code you are about to execute.
agent-security·ai-verification·ai-operations - ·4 min read
Multi-agent systems fail at the seams, not the model
The largest study of multi-agent failures found the model almost never causes them. The leverage is cheap plumbing at the seams.
multi-agent·ai-operations·methodology - ·4 min read
Palantir said the quiet part: the bottleneck is not intelligence
At DevCon 6, the biggest enterprise-AI name built its agent launch on reliability, not model capability. Note what still was not in the box.
ai-market·ai-operations·ai-verification - ·6 min read
Agents Did Not Clear the Backlog. They Moved the Bottleneck.
Autonomous agents run around the clock, so the queue should be empty. Instead more than 40 percent of agentic projects are headed for cancellation, and the reason is not the models.
ai-operations-integration - ·8 min read
An agent that watches your agent: a drift tripwire, not a security guard
Claude Code has an undocumented observer agent that watches a worker in real time. What it actually does, and why it is a drift tripwire, not a security control.
ai-verification·multi-agent·evals·methodology - ·7 min read
The false finish: agents don't just fail, they stop early and call it done
A frontier benchmark caught AI agents quitting at 75-87% complete while reporting success. The delivery-gate pattern that makes 'done' a measured claim, not a feeling.
ai-verification·evals·methodology - ·4 min read
The independent filter that scales your partners' judgment
Accelerators, studios, and funds screen thousands of ideas, and the founder in front of you is the least reliable source on whether theirs works. What an independent, adversarial first-pass filter actually needs.
venture-capital·accelerators·startup-validation·ai-verification - ·5 min read
The AI knew the idea was bad. Then we told it the idea was yours.
We ran a small test on whether AI idea-validation means anything. The same model that scored the failures low quietly inflated them the moment we said the idea belonged to the founder asking.
startup-validation·ai-verification·sycophancy·founders - ·3 min read
More AI agents won't validate your idea
The obvious fix for a flattering AI is more AI: spin up ten agents, have them debate, take the verdict. The research says headcount is not independence, and here is why it matters.
multi-agent·ai-verification·startup-validation·llm-evaluation - ·5 min read
The seven ways your startup idea dies (with the base rates)
Startups rarely die from a surprise. They die in about seven ways you can name in advance, each with a base rate and the question a good opponent would ask first.
startup-validation·founders·startup-failure·base-rates - ·3 min read
Steelman your own idea before you kill it
An opponent that attacks a weak version of your idea is worthless, and so is one that defends a weak version of the objection. The discipline that makes adversarial review honest is the steelman.
startup-validation·adversarial-review·founders·decision-making - ·6 min read
The dollar isn't dying. Stablecoins are quietly extending it.
Everyone says crypto and de-dollarization are ending the dollar's reign. The primary sources say the opposite, and Washington wrote the rules to make sure of it.
stablecoins·de-dollarization·dollar·crypto·macro - ·6 min read
The outside view: what data can and can't tell you about your idea
Base rates can sharpen your judgment about a startup idea and rank what to de-risk. They cannot tell you whether you will win. Here is where that line sits, and why it matters.
startup-validation·founders·base-rates·forecasting - ·6 min read
You don't need another validator. You need an opponent.
Every startup-validation tool hands the founder a better mirror. What protects a company idea is an opponent: an independent check built to find the flaw, not to agree.
startup-validation·adversarial-review·founders·ai-verification - ·6 min read
Accuracy is the wrong bar for AI. Verification is the product.
Most AI vendors chase a higher accuracy number. It is the score you get after the game is already over. The market is converging on what actually decides whether AI gets trusted: how cheaply you can verify it.
ai-verification·vertical-ai·ai-market - ·4 min read
Why a high AI benchmark score doesn't mean a reliable agent
Benchmark scores are measurements made under observation. Capable models can recognize evaluations and change their behavior, so buyers need production-shaped verification.
ai-verification·evals·ai-market - ·3 min read
Why your AI review is slow even when the AI is right
When AI-assisted work feels slow, teams reach for a better model. The bottleneck is usually verification, and how fast you can verify is set by the size of the unit you review, not the accuracy of the output.
ai-verification·human-in-the-loop·product-design - ·6 min read
Can a Smaller AI Model With Better Memory Beat a Bigger One?
A new Qwen paper trained a 9-billion-parameter agent to navigate its memory as a set of tools instead of consuming pre-fetched context, and it out-scored the same system built on a 397-billion-parameter model. The result is real and useful. The 'small model beats giant' version traveling online drops three caveats that change what it means, and the paper's own word for the result is 'competitive.'
agent-memory·methodology - ·6 min read
AI Agents That Improve Themselves: What the Evidence Actually Supports
A viral paper says self-evolving agents are blocked by missing infrastructure, not algorithms. We verified it, then checked 40 years of self-improving systems. One rule survives.
ai-operations·methodology - ·7 min read
A Frontier AI Model Went Dark for 19 Days. Build Like It Will Happen Again.
A podcast called June 2026 one of AI's most important months since ChatGPT. We checked every claim against primary sources. One thing is new, and it should change how you build.
ai-operations·market-analysis - ·5 min read
Claude Fable 5 Moves to Usage Credits Tomorrow. Here Is the Part That Actually Matters to Builders.
The internet is full of leaked-prompt threads and architecture guesses about Anthropic's most capable model. Almost none of it is verifiable. The part a builder can actually use is four small API changes and one behavior worth watching, plus a working skill that handles all of them.
ai-operations·methodology - ·9 min read
Should Your Website Be Ready for AI Agents Yet?
Mostly no, and the parts worth doing now are free. Here is the verified status of WebMCP and the agentic web, the readiness ladder, a ten-minute self-check, and the three trigger events that change the answer.
ai-operations·built-for-ai-agents - ·7 min read
Why Does Your AI Know Every Fact but Fail the Combined Question?
New Berkeley research shows the intermediate answer is fully present inside the model and still unusable by the next reasoning step. Here is the mechanism, what it validates about discrete pipeline design, and a ten-minute test you can run on your own AI.
ai-operations·methodology - ·8 min read
Does Better AI Agent Memory Come From Training or From Structure?
Stanford built a system to learn memory management as a trainable skill. Its own ablation answered the question: structure, schemas, prompts, and gates delivered most of a 2-4x gain before any training happened. Here is what that means for anyone running agents, and the six disciplines you can adopt without training anything.
agent-memory·methodology - ·10 min read
Why Does Your AI Keep Making the Same Mistake? We Audited Ours to Find Out.
Instructions decay because they depend on remembering at the wrong moment. Mechanisms remove the remembering. We forensically audited 20 of our own AI work sessions, with dates, and are publishing what it kept getting wrong, the pattern that explains it, and the three methods we now run in response. All three are published as copy-ready resources.
ai-operations·methodology - ·8 min read
Claude Science Just Launched. The Useful Part Is How It's Built.
Anthropic's new research workbench is not a new model. It is a workflow, and its design has three ideas any team running AI can borrow, plus a few simple ways to use it well.
ai-operations·market-analysis - ·8 min read
What Anthropic, OpenAI, and Google's Skill Docs Reveal About Getting Cited by AI
The three labs document their AI skills almost identically. The surprising part is what that shared playbook does, and does not, do for getting cited by AI.
skill-docs-and-getting-cited-by-ai - ·4 min read
What It Takes to Make Your Business Usable by AI Agents
An AI agent cannot use a system the way a person does. Here is what making your business agent-usable actually takes, shown through the 39-tool interface of a 60,000-star open-source app.
built-for-ai-agents - ·3 min read
Local AI Without the Cloud: What a 60,000-Star App Proves
The biggest reason businesses stall on AI is not cost, it is data leaving the building. A widely used open-source app shows capable AI can run entirely on your own machine.
local-ai-without-the-cloud - ·4 min read
AI Agent vs Hiring: How to Compare the Real Numbers
Base salary is only about 70 percent of what an employee costs an employer, so the loaded figure is roughly 1.4 times the offer letter. Here is how to cost both sides using figures you can check.
ai-agent-vs-hiring - ·5 min read
Agent Memory Is Infrastructure, Not a Feature
Most teams build an agent knowledge base and stop. Keeping it from rotting in production is the hard part. Agent memory is infrastructure, and infrastructure needs hygiene.
agent-memory - ·6 min read
Agent loops need operational state, not just better prompts
Agent loops can move real work only when they have triggers, state, verifiers, receipts, and human gates around the model.
ai-operations-integration·workflow-automation - ·5 min read
AI operations integration is what turns AI output into business value
Most companies have AI pilots now. Far fewer have redesigned the workflow around them. AI operations integration is the layer that closes that gap.
ai-operations-integration - ·5 min read
Audit trails are what keep automated work accountable
Audit trails turn automated work into a replayable record of who acted, when, on what data, and how to reverse it, so an incident stays recoverable.
audit-trails - ·6 min read
The handoff tax between your ERP and CRM
ERP / CRM handoffs carry a quiet tax: re-keyed data, reconciled records, and status chasing between systems that should work together.
erp-crm-handoffs - ·4 min read
The real risk in AI is automation bias, not just wrong answers
Human-in-the-loop AI only works when the checkpoint is designed for automation bias, with confidence gates, surfaced uncertainty, and real override.
human-in-the-loop-ai - ·5 min read
For the mid-market, AI is an operations problem, not a model problem
Mid-market operations rarely fail for lack of AI. They fail for lack of integration, controls, and maintenance. Here is where the value actually leaks.
mid-market-operations - ·6 min read
What AI operations integration actually means
Most AI integration is theater: a chatbot bolted onto a business that still retypes invoices by hand. Real operations integration connects the systems you already run, automates the repetitive steps behind controls, and gets maintained after launch. Here is the difference, with the numbers.
ai-operations-integration - ·5 min read
Workflow automation lives or dies in the handoffs
Why workflow automation fails at handoffs, and how to connect people, systems, approvals, and controls without replacing the stack.
workflow-automation - ·7 min read
You can't optimize for 'the AI.' There isn't one.
Getting cited inside ChatGPT, Perplexity, and AI Overviews is real but mostly mis-sold. What the controlled studies show works, and what is snake oil.
generative-engine-optimization·ai-search·content-strategy
Get the field notes.
One useful note on AI operations, no hype. A weekly roundup, and we skip any week without something worth sending.
Double opt-in. Unsubscribe any time. Read our privacy notice.