Toby Fleming

When (Not If) Containers Misbehave I just published my second article, "Fargate + Lambda are better together". Running containers in production, I encountered 14 different scenarios where things got interesting. https://lnkd.in/dXnYDqv6 #AWS #Containers #Serverless #CloudArchitecture #DevOps #Lambda #Fargate

18

3 Comments

Chunking and snippet extraction has been a huge focus lately - my latest blog dives into some of the work we've done on this to date, including support for a chunk rescorer in our semantic reranking retriever, as well as some useful ES|QL primitives to get more visibility into chunks and snippets. #elasticsearch #semanticreranking #snippets #chunks #chunking #esql https://lnkd.in/e6UD7iii

13

Seeing quite a few discussions lately about Kafka/Iceberg integrations being "zero-copy" or not. I think this is largely missing the point. First, where I agree is that this integration should be "zero-effort" for users. Materializing a Kafka topic into an Iceberg table shouldn't require more than a click of a button. Queries should provide a uniform way for accessing the data in both a topic and the corresponding table. This is the stream/table duality, and it should Just Work™. Now, whether this requires to store the bytes of data once in a Kafka topic, and a second time elsewhere for table access, shouldn't really matter from a user perspective. I'd argue storing the data twice is actually a benefit, and in fact it's a pattern well established: it resembles the design of WAL and table files known from databases for decades. I don't think anyone ever complained about this structure in their RDBMS? Which makes sense, it's an implementation detail, opaque to users. But as it turns out, having log and table data separately is even more advantageous for the deconstructed database that is Kafka and Iceberg: you can have multiple readers of the same log (Kafka topic), materializing views in multiple destinations and systems optimized for specific use cases. Maybe multiple Iceberg tables with different projections (think PII), maybe an Iceberg table and a full-text index in Elasticsearch, maybe an... you catch my drift. Furthermore, the log is replayable, so you can recreate views if needed, or you can implement new use cases you didn't originally have in mind. All in all, I think "zero copy" is mostly a red herring. Sure, it can be an optimization for certain scenarios, but mostly it's a distraction from the immense value you get from combining Kafka and Iceberg when done the right (seamless) way.

127

26 Comments

I've been leaning pretty heavily into AI agentic coding the last few weeks, basically a deep tire-kicking exercise trying to find the boundary of what works and what doesn't. So far it's been "breathtaking" (in the Seinfeld sense) to see what can be done with prompting, as well as how dramatically it can crash and burn. Overall I'm feeling like I wouldn't advise SW engineering as a career for someone entering the workforce, at least not the way I was taught it. Questions for folks I've worked/coded with over the last couple of decades: - Are you finding AI to be force multiplier, a distraction, or both (depending on context)? - Are you using AI to refactor/debug/improve existing code, or to write new code? - Any interesting stories to share? My interesting "mind blown by AI" was that I used Copilot to help debug an obscure issue in some 10 year old code running on an equally old Linux kernel. Not only did it teach me thing or 2 or 3 about how to configure an obsolete kernel, it also did a darn good job at finding and fixing some device driver bugs. So it ain't just for vibe coding web apps...

72

63 Comments

Duplication avoidance (zero copy) is just one aspect of "trade off optimization" in data system architecture. Any data system design must balance cost, with performance and complexity, against a number of constraints. So while there are benefits to "zero copy", it is one aspect of a much larger architectural discussion. For example, we may choose to one or more partial copies for indexes to boost read performance, we may choose to store data in both row and columnar formats. We may store precomputed data and raw data. The list goes on. So I kind of agree that zero-copy itself being a major focus can be too limiting, simplistic and an artificial battleground.

41

4 Comments

🧠 How does an intelligent agent remember what matters most, at the right time? Here’s a sneak peek into a layered memory architecture I've been working on—designed to optimize how agents handle hot, warm, and cold memory using Redis, Qdrant, and PostgreSQL. 🔥🟡❄️ From real-time context to persistent profiles, this design aims to bring speed, semantics, and structure together in one cohesive memory model. And yes—it’s not just conceptual; it’s already in implementation! ⚙️🤖 📌 Stay tuned for my next blog where I’ll break this down and show how to plug it into modern agent frameworks. Exciting stuff ahead! #AI #GenAI #RAG #AgenticRAG #AIAgents

38

Classic SRE is the only thing keeping Generative AI from being a dangerous toy. Companies are currently flooded with "AI Engineers" who can build cool prototypes but have no idea how to run them at scale, secure them, or control their costs. They are building "probabilistic software" (AI) using "deterministic ops" (traditional DevOps), and it's failing in production. Anyone agree with Gemini's take? #sre #devops #gemini

63

12 Comments

POV: It’s 2026 and your manual infrastructure is a ticking time bomb. 💣 If you’re still running terraform apply from your local terminal, we need to talk. Production-grade IaC isn't just about "making it work"—it’s about making it unbreakable. The 2026 Playbook for Terraform Elites: 🔒 State Isolation is King: Local state is a myth. Use S3/GCS with DynamoDB locking. Separate your Dev/Prod states or prepare for a $100k "oopsie." 📦 Modular or Bust: Stop copy-pasting code. Build versioned modules. Pin everything. If a provider update can break your stack, your stack is already broken. 🛡️ Zero-Trust Secrets: Hardcoding passwords in .tfvars? Straight to jail. 🚩 Fetch them dynamically from HashiCorp Vault or AWS Secrets Manager. 🤖 Policy-as-Code (PaC): Don't just plan. Use Open Policy Agent (OPA) to auto-kill any PR that tries to open Port 22 to the world. Stop clicking buttons. Start scaling logic. 🚀 What’s your #1 Terraform "lesson learned" the hard way? Drop it below. 👇 #DevOps #Terraform #CloudNative #SRE #InfrastructureAsCode

8

Colin is not only a prolific contributor to the Linux kernel but also the maintainer of stress-ng, one of the most widely used tools for flushing out bugs in the kernel, firmware, and even hardware. If you’ve ever considered supporting some good work in free software, you should take a look!

18

1 Comment

I've always found myself switching to Claude models when working through coding problems. There's just something about how Claude handles context and works its way through large code bases and complex logic. With Claude Code, that workflow has become ridiculously smooth.                                                    Also been experimenting with OpenAI's Codex (GPT-5 powered) and the results are way better than previous versions. The code generation quality has taken a real jump. Now here's where it gets interesting. You keep hearing stories about developers opening multiple coding agents, feeding them the same task, and manually cross-referencing outputs to find the best solution. Turns out there's an easier way. By integrating Codex as a skill inside Claude Code, you get insights from both agents in a single terminal session when you need. No tab switching, no copy-pasting, no manual comparison. Claude Code orchestrates the workflow while Codex brings its strengths. Best of both worlds in one window. Claude OpenAI

35

I'm sure the view I get through linked in is heavily biased by my skewer network and the algorithm, but... As far as I can tell there is a big spike in demand for engineers with 8-15 years experience at the senior/staff borderline, and a drop in demand for junior engineers. I'm assuming this is because AI tools are making the senior engineers much more efficient. Small jobs that would have been given to an intern or new grad can now be done swiftly by AI with less supervision than a human would need. As AI first senior developers are doubling, quadrupling or more their efficiency the economically valuable work that would have been done by junior engineers is disappearing. Am I imagining this trend or are others seeing it too? What longer term impacts will this have on the job market?

7

12 Comments

I 100% support this message. Corporate users should all be using an internal permanent caching mirror for both continuity of business as well as potentially security scanning up front. Companies that consume through a permanent caching mirror (i.e. it stores anything it has fetched at least once) should not be the overconsumers that Brian and others are complaining about. Another way to reduce the load would be for the IaaS providers to host their own caching proxies and require their customers to pay if they connect to the artifact storage tiers directly. But the biggest point is that it's in your interest to use a caching proxy. The companies who are thanklessly hosting terrabytes of artifacts for consumption by the community are doing so because it's the right thing... your side of that bargain is to do the right thing yourself and consume responsibly... I suspect that in the majority of cases it's that you have a CI job... if you have it pull the dependencies through a cache then you are consuming responsibly... oh and there's a second win for you too, your jobs are now faster

4

JSON-structure is a modern practical alternative or replacement for JSON-Schema by Clemens Vasters - it's a draft IETF standard and the SDK now includes a Perl implementation after some not so gentle nudging. Clemens has been enthusiastically responsive to feedback and knows his stuff, if you're working with JSON and Perl (or other languages) this is well worth taking for a spin and sending any bug reports or suggestions.. code is at https://lnkd.in/e5fZvw_F and I'm looking forward to seeing the perl library on cpan shortly. #JSON #perl #javascript

8

🦀Rust🦀 + Local LLMs, could there be a better combination? #FoundryLocal makes it trivial to code against local LLMs, including #Rust now with the new foundry-local crate I just published! https://lnkd.in/gsKjtee7

8

I spent last week at GitHub Universe 2025, soaking in the sun and the AI hype. What stuck with me wasn’t a feature or demo. It was the gap between the people building with AI and those just starting to explore it. I wrote about what that divide looks like — and why the shape of the internet might depend on how we teach the next generation to use these tools. 👉 https://lnkd.in/gwayrhzf

11