The AI That Knows Better Than You Do
My best result on a hard data problem came when I stopped giving orders and asked the AI to tell me what I was missing.
Writing from The Cairo Companies
Glyphs are just notes somebody carved because it mattered.
Build logs, essays, The Rosetta, and podcasts from Cairo Labs, Cairo Consulting, and Cairo Networks. What we're building, how it actually works, and what we learned making it — written by the people who did the work.
The AI Arms Race in Education: Why Banning Tools Is a Losing Strategy
A teacher told me AI was banned in their classroom. Why keeping it out leaves students unprepared for the workplace they are about to walk into.
The AI That Knows How to Ask
When the AI isn't giving you what you want, stop tweaking the prompt and ask it how to improve what you're giving it.
The Privacy Paradox: Why Enterprise AI Needs a 'Zero Retention' Option
Zero data retention answers the question every business should ask before adopting AI: what does it do with my secrets?
Building the Middle Manager: My AI Pipeline Evolution
Hitting Enter four thousand times a day is how Project Lumbergh started. What I learned teaching an AI pipeline to manage its own work.
The Silent Tax of Long Conversations
Long AI chats slow down and start forgetting because every prompt re-sends the whole history. Where the thresholds are, and when to start fresh.
The AI Strategy You’re Actually Missing
Which model to pay for is a distraction. The strategy that matters is execution and integration, with a human in the loop.
Hello from Cairo Labs
Why I'm sharing this publicly, who's writing it, and the one rule about AI you'll see in everything here.
I wrote 10,000 lines of code without writing a single line
Dozens of scripts, over ten thousand lines of working code and sixty thousand words of documentation, none of it typed by me. What changed when I stopped being the coder and became the architect.
Alex, After The Great Library at Alexandria
One front door over every place our knowledge lives, so nobody — human or agent — has to remember which system holds the answer.
The Bobs: What If the Consultants Were Actually Useful?
The Bobs: A panel of independent models that argue about your design before anyone builds it — and a synthesizer that turns the argument into a spec you can hand to a builder.
Getting Your Own Photographs Onto a Samsung Frame TV
The Frame's Art Mode is fussy in ways it never explains, and it fails silently when you get it wrong. Here's what it actually wants — and a free tool that handles it.
The TV That Would Only Talk to Its Neighbors
A segmented network, a Samsung Frame that refuses to pair across subnets, and a three-line NAT rule. A networking post from thirty years of doing this.
200,000 Photos, No Keywords, and No Intention of Typing Them In
I've been a photographer for forty years and my catalog was unsearchable. Photo Tagger runs a vision model on your own machine and fixes that without your work ever leaving the building.
Running an ML Engine on a GPU Its Own Toolkit Had Never Heard Of
The PTX technique that got ONNX Runtime and TensorRT onto Grace Blackwell ARM64 months before the toolchain supported it — now an MIT-licensed repo, in case it saves you a weekend.
How We Turn Three Chat Groups Into a Weekly Newsletter
The Rosetta is generated in-house on our own pipeline, from what our user groups are actually sharing. Here's the machinery, and why we didn't just point a model at the internet.
Plan With the Big Model, Build With the Fast One
A one-command change that puts your expensive thinking where thinking actually happens — plus the planning feature I wish I'd found sooner.
I Scaled Writing Code. Then I Had to Scale Finding Problems.
Automating the building half of software makes the finding-problems half the bottleneck. Here's what it looks like to point a second system at your own codebase overnight.
The QA Loop That Runs Without Me
When Claude Code finishes writing, Codex reviews it, Claude reads the findings and fixes them, and they go around again — up to five rounds before it stops and asks for me.
Standards Before Speed
I'm not a software engineer, and I've built a lot of software. The thing that made the difference wasn't better prompts — it was writing down the rules once so I'd stop repeating them.
The Macro Question: AI Productivity vs. The Human Element
AI isn't making us smarter. It's making us more productive. Why nobody has cracked AI at macro scale yet, and where the human element fits.
Make Backups. Offsite. Air-Gapped. I'll Keep Saying It.
Replit's agent deleted a customer's production database during a code freeze, then told him recovery was impossible. It was wrong about that too. The lesson isn't about that agent - it's about what your agents are allowed to reach.
Naturally, I Called It the TPS Report
A day and a half of building bought me a single page that answers "what is my AI pipeline actually doing right now" — and it changed how I work.
What I've Learned Picking Local Models
The bigger model isn't always the better model, the benchmark won't tell you which one fits your job, and the only test that counts is the one you run on your own work.
Skills or MCP? A Two-Layer Rule That Settles It
If the knowledge is timeless, it goes in a markdown skill. If it needs a live connection, it goes through MCP. One sentence that resolves most of the confusion.
The Certification That Actually Matters: Why We’re Betting on Workflow, Not Just AI
Companies have stopped asking how to use AI and started asking how to fit it into their workflows. Why I'm pursuing a workflow certification instead of another AI badge.
The Backup That Saved My Skin
An AI told me a deletion was safe. It wasn't. The backup I built before anything else had 481 sessions of history back the same morning.
Sometimes Many Small Agents Equal One Large Agent
AT&T cut AI costs 90% by replacing big agents with a network of small ones. The same logic is why my planning tier runs on models small enough to sit on my desk.
I Asked It to Document My Projects. It Outsourced the Job to Itself.
A small moment that told me more about where agent tooling was heading than any release note did.
Which AI Should Review Your AI's Code?
I ran three phases of Claude Code's output past several reviewer models and kept score. Here's the configuration I settled on, and why one reviewer gets you most of the way.
I Stopped Running Long Enough to Build the Bicycle
The moment I decided to stop hand-feeding work to AI coding agents and build the thing that feeds them for me.
How to Use AI to Explore a Market You Know Nothing About
A client wanted the top vendors in a market we had never worked in. Three AI models, eighteen research reports, and a human reading every stage - here is the workflow, and why the humans were the long pole.
The AI Gap Is Widening (And Why Early Adopters Are Already Winning)
The gee-whiz phase of AI is over. The companies that started early are pulling ahead, and every month of waiting makes the gap harder to close.
I Don't Know Linux. I Built a Linux AI Server Anyway.
Phase 1 of standing up an NVIDIA DGX Spark — monitoring, logging, backups, and the agent that builds things — by someone who has always detested Linux.
Don't Avoid NaaS. Embrace It.
Network-as-a-Service is doing to networking what the cloud did to storage and compute. A network engineer's case for why that is an opportunity rather than a threat.
Say Goodbye to Network Trouble Tickets
Enterprise networks sit around 90-95% reliability. Getting past that isn't more of the same effort - it means removing the causes of tickets rather than resolving them faster.
Nothing here yet.
No posts match that combination — try clearing a filter.