iiinova — where your digital self keeps evolving

"Million" sounds like a big number, but in context, it's really not. The water cycle also means water doesn't just disappear when it evaporates.

100

Millions of gallons across the entire industry, yes it is. A single paper mill uses 17,000 gallons of water per ton of paper produced, and add loads of pollutants to it. Ag measures their water use in acre-feet, which is the amount of water needed to flood an acre of land by x feet.

200

Maybe We Shouldn't Be Reviewing All This Code

At @martinfowler.com's site, a great piece by Rachel Laycock on how agentic development has changed the code review process and how we can get the things we used code review for (hint: not finding bugs) by rethinking our process

martinfowler.com

Or, perhaps the problem isn't that AI has broken code review, maybe it’s that we've been using code review to solve the wrong problems

120

Kyle @kylestratis.com · 27/08/2026

It’s more that they’d rather not pay for the water over time. The water usage is laughably small compared to most other industrial uses, that doesn’t mean they’re not going to try and minimize costs.

100

MCP just had its biggest rewrite. I joined ODSC’s AI X Podcast to talk through statelessness, designing tools agents can actually use, agent skills, what the changes mean for people building with the protocol, and how fun it is writing a book on a rapidly changing protocol. Listen here:

linktr.ee

ODSC AI Official Music - Listen to songs on Soundcloud | Check for Videos on Linktree

View odsc’s Linktree to discover and stream music from top platforms like Soundcloud here. Your next favorite track is just a click away!

210

Kyle @kylestratis.com · 20/08/2026

My recharge time is ending, which means my books are opening back up: 15-20 hours a week this fall. Agent builds, the evals that show you where they break, cloud infrastructure, and custom MCP workshops. Message me or book a consult: stratisdatalabs.com

stratisdatalabs.com

Stratis Data Labs

Stratis Data Labs is an AI, data engineering, and cloud engineering consultancy

000

Kyle @kylestratis.com · 12/08/2026

Hell yeah! Thank you

010

Kyle @kylestratis.com · 12/08/2026

Now that the book is mostly done, The Signal Path is back! In it, we talk about the new awkward teenage phase of agents: MCP’s enterprise-focused overhaul, @scoiattolo.mountainherder.xyz’s article on best practices on using agents, and stories of agents getting too big for their sandboxes.

thesignalpath.xyz

Coming Up for Air

It's been a while, hasn't it? I've been incredibly busy, juggling my part-time work, a string of big enterprise contracts (Koch's Principle Companion, a new learning agent for Microsoft, and an MCP wo...

040

Python SDK v2 is a matching rework: MCPServer replaces FastMCP, fields are now snake_case, and context is injected explicitly. It's a breaking change, but a lot of care has been taken to ensure backwards compatibility, and setting mode="auto" when making a connection will handle everything for you.

100

Also new: MRTR replaces server-initiated requests (supporting statelessness), an extensions framework lets features grow outside the core release cadence, tasks are now an official extension, and roots, sampling, and logging are deprecated with clear migration paths. Plus OAuth 2.1 across the board.

110

The biggest change: the protocol core is now stateless. The initialize handshake and protocol-level sessions are gone, and every request carries its version, client info, and capabilities in _meta. This makes scaling your remote servers much easier.

100

The MCP 2026-07-28 spec is final today, and v2 of the Python SDK has shipped with it. Congrats to the spec maintainers, SEP authors, and SDK teams - so much hard work went into maturing the protocol, and the SDK folks ran an epic marathon to keep everything in sync with the spec.

120

dan @danabra.mov · 30/06/2026

Will do, love the idea!

020

Reposted by Kyle

i want to spread the word about atproto. if you’re running a podcast, i’d be happy to hop in to talk about it. i understand it decently well and can explain it without much jargon. it’s one of the most interesting developments in the last 30 years of web, and i can defend that. please dm to set up!

3745455

Kyle @kylestratis.com · 03/06/2026

Thanks to @sequoia.pub my blog is now on the ATProto! Do I do the Ghost newsletter next?

030

Kyle @kylestratis.com · 19/05/2026

One of the things I've been heads down on lately has been this AI companion for the book Becoming a Principle-Driven Leader by Charles and Chase Koch. I got to work on some interesting problems, develop techniques based on modern andragogy research, and deploy a multiagent system to the public.

linkedin.com

Chase Koch on the Principles Companion App that Tribe AI Built | Tribe AI

Most AI is being built to think for you. This one was built to elevate your thinking. Koch, Inc. asked Tribe a different question than the one most companies are asking: could AI develop human judgme...

000

Kyle @kylestratis.com · 02/05/2026

I should be more active on this account, maybe when I’m not spending all my waking hours writing

010

Kyle @kylestratis.com · 02/05/2026

If you have access to O’Reilly’s library, my MCP book (AI Agents with MCP) is in early access and has a section on testing servers. Whether I’m making tools available with MCP or otherwise, I generally like to use Promptfoo to do this kind of testing, but you’ll have to build a small client.

000

Kyle @kylestratis.com · 03/03/2026

Security has been a hot topic in the MCP world since its launch, and this chapter aims to consolidate all the current best practices on the topic in one place. If you subscribe to O'Reilly's learning platform, you can read it now! learning.oreilly.com/library/vie...

oreilly.com

AI Agents with MCP

Since its release in late 2024, Anthropic's Model Context Protocol (MCP) has redefined how developers build and connect AI agents to tools, data, and each other. AI Agents with MCP... - Selection from AI Agents with MCP [Book]

010

Kyle @kylestratis.com · 03/03/2026

What I think is the most important chapter of AI Agents with Model Context Protocol (MCP), and most difficult to research and write, is now available in early release. In chapter 7, you'll learn how to test, secure, and distribute your MCP servers.

111

Kyle @kylestratis.com · 25/02/2026

I have (nearly) a whole chapter about it in my book and I still don't think I did it justice

120

* nearly every reward schedule is intermittent

000

I lean towards the latter though because in the literature it’s “intermittent reinforcement” not reward, and IR is a very large class* of reward schedules that have pretty different behavioral effects and consequences. Slot machines employ variable reinforcement, a specific intermittent schedule.

100

Mostly because it’s so far off from an actual intermittent reward that to call it such is to completely misunderstand either LLMs or Skinnerian behaviorism. Or I guess both is also a possibility.

100