Pure Signal AI Intelligence
Here's the tension at the heart of enterprise AI right now: most companies are bolting AI onto existing products and calling it a strategy. A few are asking a much harder question — what does it mean to actually own your intelligence?
The Agentic Workforce Is Here — And It's Reshaping Risk
DJ Sampath, Cisco's head of AI software and platform, is making a striking prediction. Within twelve months, he expects AI agents to autonomously resolve roughly eighty percent of routine network incidents — pattern-based outages, standard remediations, the predictable and procedural. The remaining twenty percent — multi-vendor, legacy-heavy edge cases — will take longer. But as he puts it, progress will compound.
This isn't just an operational claim. It's a fundamental reshaping of how teams work. Every leader, Sampath argues, will soon manage a constellation of agents working in parallel — investigating, analyzing, executing — while humans move up the stack to creativity, judgment, and strategic direction. The premium shifts. Routine thinking gets absorbed. What remains is the work that actually requires being human.
Here's what's interesting, though. The more autonomous these agents become, the more dangerous they get — not just powerful. Sampath identifies agent compromise as the most urgent AI security risk right now. As agentic systems gain access to data, invoke tools, and make decisions independently, they become a new attack surface. They can be hijacked, impersonated, or manipulated to exfiltrate data at machine speed.
The connective tissue is the vulnerability. Protocols like Model Context Protocol — the standard linking agents to tools, data, and each other — scaled faster than the security around them. Sampath's prescription is clear: anything involving trust, access, or irreversible actions should never run fully autonomous. The right model isn't humans out of the loop. It's AI in the loop — agents handling routine and low-risk work at speed, humans holding authority where consequences are real.
Owned Intelligence vs. Rented Intelligence
This brings us to the deeper strategic question Sampath is raising — one every company building on top of foundation models should sit with.
Adding a generative API to an existing product isn't a strategy. It's a feature. The companies Sampath describes as "a thin shim on top of a model" — their days, he says bluntly, are numbered. Sustainable advantage comes when intelligence is embedded into the product itself. When models are trained on proprietary enterprise data, they improve continuously. The product becomes the model. The model becomes the product. That closed loop — fueled by data no one else has — is the actual moat.
Sampath goes further. His personal conviction is that the future shouldn't belong to a handful of centralized AI providers. Intelligence, he argues, should be owned by enterprises — and ultimately by individuals. That means building the full stack: developing, fine-tuning, deploying, and governing models on your own terms.
It's a provocative thesis when most enterprises currently rent their intelligence from OpenAI, Anthropic, or Google. And it raises a question worth sitting with: are you building AI capabilities that compound over time — or renting them from someone who could change the terms tomorrow?
The Practical Workflow Behind the Philosophy
Sampath also shared how he actually works — and the details are worth noting. He separates idea generation from evaluation, drafting strategy memos in one model, then bringing them into a second model to critique and improve. He uses Cursor — a code editor with deep AI integration — to store context in markdown files, building what he calls a long-term thought partner that understands his frameworks and past work over time.
This is the compounding system idea made concrete. One-off AI interactions are low-leverage. Structured, persistent, context-rich workflows are where the real productivity gains live.
The throughline across all of this is compounding. Agents that get better as they handle more incidents. Models that improve as they ingest more proprietary data. Workflows that accumulate context over time. The organizations that figure out how to build these feedback loops — not just deploy AI features — are the ones Sampath expects to define the next era of operational performance.
The question isn't whether your company uses AI. It's whether the AI you're using is making you smarter over time, or just making someone else's model better.
HN Signal Hacker News
☀️ Morning Digest — Hacker News for the Curious
Your friendly guide to what's happening in tech today
🔝 Top Signal
[Japanese Woodblock Print Search: A Beautiful Corner of the Internet You Didn't Know You Needed](https://ukiyo-e.org/)
A searchable database of thousands of Japanese woodblock prints — called ukiyo-e — pulled together from museums and universities around the world. What makes this extra delightful is who built it: John Resig, the same person who created jQuery (a hugely influential programming tool from the 2000s that made it much easier to build interactive websites). He built this entirely out of personal passion for the art form, using computer vision (software that can "see" and compare images) to group similar prints together across different collections. The thread is warm and appreciative — one user, chickensong, even used it to identify prints on boxes their grandparents brought back from Japan. Bonus: someone linked to a [Twitch stream](https://www.twitch.tv/japaneseprintmaking) of someone doing woodblock printing live. A rare feel-good corner of HN.
[HN Discussion](https://news.ycombinator.com/item?id=47107781)
[Back to FreeBSD: Part 1 — One Developer's Case for Ditching Linux](https://hypha.pub/back-to-freebsd-part-1)
FreeBSD is an alternative operating system (like Linux, but a separate family tree) that's been around since the early '90s. It's less common than Linux today, but has a devoted following among developers who appreciate its clean design. This article makes the case for switching back to FreeBSD — and one of its main arguments is about jails vs. containers. A quick explainer: containers (like Docker) are a popular way to package software so it runs the same way on any computer, like shipping your app in a standardized box. Jails are FreeBSD's older, simpler version of the same idea. The author argues Docker became a bloated, overcomplicated mess. The HN community pushes back thoughtfully — commenter matheus-rr makes the excellent point that Docker didn't win because of its technical isolation, but because of its ecosystem: a public library of ready-to-use software packages and easy-to-share setup files. Jails were technically solid first, but harder to get started with.
[HN Discussion](https://news.ycombinator.com/item?id=47108989)
👀 Worth Your Attention
[What's the Best Way to Learn a New Language?](https://www.bbc.com/future/article/20260220-whats-the-best-way-to-learn-a-new-language)
A BBC piece on language learning that sparked a fun side conversation on HN: what's the best way to learn a new programming language? The consensus from commenters: build something you actually need, not another to-do list app. Commenter dabinat suggested porting a real tool you've already built into the new language — you already know what it should do, so you can focus on learning the new syntax. Good practical advice for anyone dabbling in coding.
[HN Discussion](https://news.ycombinator.com/item?id=47108977)
[Show HN: Elecxzy — A "Lightweight, Lisp-Free" Emacs Editor Built on Electron](https://github.com/kurouna/elecxzy)
This one generated more heat than light, but it's a fun read. Emacs is a legendary, highly customizable text editor beloved by programmers — its superpower is that you can reprogram it while it's running, using a language called Lisp. This project tries to keep Emacs's keyboard shortcuts while ditching Lisp... and replacing it with Electron (a framework notorious for using lots of memory). The community was gently brutal: calling something "lightweight" while using Electron is a bit like advertising "low-calorie" deep-fried butter. Commenter maybewhenthesun put it well: "Lisp-free Emacs is like tomato-free ketchup." That said, writing your own editor is a classic programmer rite of passage, and several people gave the creator credit for that.
[HN Discussion](https://news.ycombinator.com/item?id=47100328)
💬 Comment Thread of the Day
From the FreeBSD discussion — why did Docker win if jails were technically better?
Commenter matheus-rr laid it out clearly:
> "The jails vs containers framing is interesting but I think it misses why Docker actually won. It wasn't the isolation tech. It was the ecosystem: Dockerfiles as executable documentation, a public registry, and compose for local dev. You could pull an image and have something running in 30 seconds without understanding anything about cgroups or namespaces. FreeBSD jails were technically solid years before Docker existed, but the onboarding story was rough."
And steve1977 got a laugh with a one-word edit:
> "You could pull an image and have something running in 30 seconds without understanding anything." (Fixed that for you 😉)
This is worth reading because it captures a recurring theme in tech: the better technology doesn't always win. What wins is the technology with the better story, easier entry point, and stronger community around it. Docker made containers approachable for millions of developers who had never heard of FreeBSD jails. That's a lesson that applies far beyond operating systems.
[HN Discussion](https://news.ycombinator.com/item?id=47108989)
⏭️ Skip List
- Back to FreeBSD: Part 1 — Worth skimming if you're a developer curious about operating systems, but if you've never set up a Linux server, the jails/containers debate will feel abstract. Come back to this one later.
- Elecxzy (the Electron Emacs) — Unless you're already an Emacs user, the comment thread is mostly inside baseball. The jokes are good but the project is early-stage.
💡 One-Liner
Today's HN reminded us that the internet's best surprises are still built by one passionate person in their spare time — like the creator of jQuery quietly building the world's best Japanese woodblock print database because he just really loves the art form.