Daniel Barker

I've authored a blog for developers who are new to vector embeddings and semantic search - if that's you, hopefully you'll find it helpful! #semanticsearch #vectorsearch #vectordatabase #vectorembeddings #densevectors #sparsevectors #elasticsearch https://lnkd.in/eAU8uACZ

36

1 Comment

"I have only made this letter longer because I have not had the time to make it shorter." - Blaise Pascal'' This also applies to software demos. We've tightened the onboarding workflow and now have a 60 second demo of the journey from a blank slate in Claude Code to a deployed preview application on VeilStream.

1

Grow engineering velocity without growing the security backlog. Build security into every commit so code ships safely, no matter whether it’s written by humans or AI.

3

Using rescoring for kNN vector searches improves search recall, but can increase latency. Learn how to reduce this impact by leveraging direct IO: https://gag.gl/m7C6fc

22

Build a multi-agent system using WRITER + Strands Agents 🚀 We recently released our WRITER Strands Agents SDK integration. It was great diving head-first into developing some of our associated developer docs, examples GitHub repository, and creating my first WRITER engineering blog post. 🥳 To show you just how powerful this combo is, here’s a quick walkthrough of a multi-agent system that: ✅ Uses WRITER’s Palmyra X5 foundation model as the central orchestrator managing a multi-agent system via Strands Agents SDK   ✅ Leverages Palmyra X5 to classify and route queries to domain-specific agents that use Palmyra Creative, Palmyra Fin, or Palmyra Med   ✅ Connects each specialized agent to prebuilt tools via Strands Agents utilities, including calculator, web_search, and file tools to support their prompted tasks Dive in, test it out, and let us know what you think!   Further reading/resources: 📚 Blog article: https://lnkd.in/gJExZdV6 📖 Getting started guide: https://lnkd.in/gbXxF4gQ 💻 Examples repo: https://lnkd.in/g8shNNKn

57

5 Comments

I'm excited to announce Planetary (https://lnkd.in/emPKrsie): a new GA4GH-compliant TES server built on top of Kubernetes and Rust. To give you a sense of the performance you can get out of Planetary in the cloud, we did a stress test of 10,000 tasks submitted to Planetary through Sprocket on Azure Kubernetes Services with auto-scaling enabled. Planetary was able to accept and successfully complete all of the tasks with about ~50% of those tasks being executed _simultaneously_. Our quotas in Azure appear to be the bottleneck, which is "good" from a perspective of the scalability of the system. Planetary is designed to work well with any execution engine that can submit TES tasks, and it's co-developed to ensure close compatibility with Sprocket. We've deployed Planetary to Azure with success, and we're looking for users to try deploying it to other clouds. At the same time, this is alpha software, so keep that in mind if you're giving it a shot. If you're interested in collaborating on this project with us, you can find us in the #sprocket channel on the OpenWDL Slack (https://lnkd.in/eP_cYTTS)!

63

5 Comments

From Claude Code for developers to Claude Cowork for everyone else—this is a real shift. Stop controlling prompts and start controlling outcomes. Instead of just chatting, you give Claude access to a local folder so it can read, edit, and create files to tackle multi-step tasks like organizing docs, building spreadsheets, or drafting reports. It’s the move from “assistant” to true AI agents at work. Sure, it has bugs—but directionally, it’s very promising. #AI #Claude #vibecoding https://lnkd.in/ghmFtTFp

6

2 Comments

As COO at www.jaiinfoway.com this is an important signal for where agentic AI is heading. Non technical implementations are catching up fast but the real story is iteration speed and learning loops. Building Cowork entirely with Claude Code shows how agents are becoming first class workers not just assistants. At www.jaiinfoway.com we see this shift clearly as enterprises move from copilots to autonomous task execution with guardrails. UX will evolve tooling will mature but the direction is set. Curious how others here see non technical users adopting agent driven workflows this year. Are we ready for agents as coworkers. #Jaiinfoway #AgenticAI #Claude #AIAgents #EnterpriseAI #FutureOfWork #AIProductivity #AITransformation #AILeadership

2

Infrastructure decisions are rarely just about technology. They are about focus. Reading about Uber’s uForwarder, their internal Kafka reliability layer, reinforced something I have seen running systems at scale. Kafka is a powerful backbone for event driven architectures, but it is not a complete reliability strategy. In revenue critical environments, guaranteeing message processing at scale takes more than publishing and consuming events. You add retry logic, temporary queues, databases to track state, and background jobs to reprocess failures. Each layer solves a real problem, but it also increases operational responsibility. Over time, teams can end up running infrastructure that is not core to the business. If building it gives you clear competitive advantage in your market, build it. If it does not, buy it. Owning infrastructure is a long term commitment. Choose carefully what you decide to own. #Kafka #EventDrivenArchitecture #SoftwareEngineering #BuildVsBuy

11

1 Comment

We just published a new post on the Tensor9 blog about a problem that comes up in almost every vendor conversation we have: what happens when your application depends on a managed service like DocumentDB, and your customer can't run on AWS. https://lnkd.in/gzamYh2d

6

Every dev embracing AI is talking about Claude Code and OpenClaw. It’s very evident to me that Claude is the current winner in agentic software development, and OpenClaw is the clear winner for task automation. The thing they have in common isn’t the model they’re using; it’s the infrastructure, the ease of use, and how turnkey the systems are. I’ve been a primary user of OpenAI, but they haven’t done much to make using their models easier to interact with. That’s why I think bringing on the OpenClaw creator is one of the smartest moves they’ve made in the last year. Most of the consulting work I do around AI is about infrastructure and implementation. Very little of it is focused on generating a company’s “own model,” beyond some minor post-training or RAG implementations. And honestly, I think this year the biggest winners in the AI space won’t be the ones producing the models it’ll be the ones making them easy to use for low-tech and no-tech users. We’re at a point where the models are good enough to handle far more cases than we’ve actually put into practice. If I was a VC that’s where I would be putting my investment.

17

4 Comments

Tired of long CI builds and rising compute costs? Coda was too. They trialed Aspect Workflows, saw a 10x improvement in build times and a 67% reduction in compute costs, and decided to roll it out more broadly. "We went from having significant limits in CI and tools to where the limits are now just due to our code." — Neal Norwitz, Software Engineer at Coda Read more about the challenges they faced and the results they saw: https://lnkd.in/eA_pwcDQ

5

Just published a new blog post on how engineering teams can “shift down” compliance into platform abstractions. I share thoughts on compliance leverage, OSCAL components vs capabilities, and how platforms can carry the heavy lifting for controls like TLS and mTLS—freeing developers from duplicating effort. Read it here: https://lnkd.in/eh5cABU5 #OSCAL #ComplianceEngineering #PlatformEngineering #ComplianceAutomation

9

Code Reviews coming to Claude! https://lnkd.in/egPBurcF I've been encouraging teams to turn on automatic PR reviews by Copilot. It often finds missed edge cases, offers refactoring opportunities, and brings a lot of value at zero marginal cost (once set up). That said, it's Copilot - pretty good but not SOTA. When I run a review manually using Codex 5.3, it tends to nail the important stuff and, unlike Copilot, does not "nitpick". Curious to test CC Reviews once they hit the Pro plans. For now, my go-to workflow is: develop using SOTA model 1 (Opus or Codex) then have the other model review. Just like with two different people doing the work vs the reviewing, my anecdotal experience is that cross-model reviews tend to find more bugs.

3

Wittling a retrospective of The Apache CMS as it related to the Apache OpenOffice project… When it’s further along I will share URL, but in the meantime if you have any anecdotal commentary to add please reach out.

This is an older talk, but one I refer to often and recommend to current tech leads and those on a path to becoming a tech lead. Eryn summarizes the shift from being a full-time coder to balancing technical oversight with advocacy, facilitation, and motivation - ensuring that everyone else has what they need to succeed. Technical leads are the bridge between the business stakeholders' goals and the engineering team's ability to deliver. Definitely worth a watch (on repeat)! #engineering #leadership #techlead #leaddev https://lnkd.in/gDypDS78

12

Yes - technically, platforms like React that imply that they simplify development can enable quite the opposite - over-engineering on a scale that would have been difficult to believe a decade or two ago. The often missed downside of these beasts is that as a team slips down the complexity slope, they find themselves needing specialists in this tooling. Which is so silly - but it’s one of the top reasons that organizations fall into the fractional stack developer antipattern. If folks are being honest with themselves - there’s not much of a case to be made to use anything more than plain JavaScript for modern web solutions. Exercising this discipline, you won’t slide toward the need for specialists - and as an added benefit you won’t pollute your customers’ browsers with hundreds of megabytes of useless nonsense.

4

Taking shortcuts is not a mistake. It is a decision. In our Django monolith, those decisions started to slow us down. Changes required too much context. Side effects became harder to predict. Small updates carried more risk than they should. This article shares how we are paying down technical debt in Django without a rewrite. https://lnkd.in/gYZNYbED

33