Dan DeLauro

However you feel about AI, it's difficult to ignore just how much better fuzzy, semantic search paired with an MCP server attached to a content database is compared to typical PHP / AJAX SQL queries. We haven't decided whether this is something that will ever end up being available to the public, but for my own uses I've completely given up using any of our native CMS tools to find things we've previously covered, and I'm using The Interline's experimental MCP server to poll our nearly-six-years of back catalogue using natural language. With all the talk about agentic AI (and we'll have an analysis of the latest in that space tomorrow), it's easy to overlook what a shift if is to be able to just talk to big databases of non-normalised content and get results that don't require you to pre-parse your language to match what a machine expects.

7

2 Comments

Your dev team isn’t slow, The problem isn’t people, it’s the pipeline. One of the most common friction points I see in scaling SaaS teams is confusion around delivery speed. Releases slow down, bugs get missed, handovers become patchy, and the immediate assumption is that the dev team isn’t fast enough. You can have strong engineers, but if staging and production aren’t in sync, if there’s no CI/CD pipeline, no automated testing, everything slows, not because of individual performance, but because the system surrounding the team is unstable. I’ve seen companies lose two-week sprints to merge conflicts, unclear rollback paths, or last-minute hotfix chaos. This isn’t about productivity, it’s about process maturity. The dev team builds. But it’s the infrastructure that determines how confidently, how fast, and how cleanly those builds get into users’ hands. When that system is scattered, even great teams feel slow. Before pointing at the output, check what it's running on. Is your delivery pipeline built for speed or friction?

8

5 Comments

I have quibbles with the N/N Group's recent information architecture change. Did they research this? Format-based navigation is not ideal for either users or searchers. Who goes to a website and says "oh wow I'd love to see some articles and videos. doesn't matter the topic or timeliness, just articles and videos"? (No one. No one does that. Format-based navigation is a massive misfire for all content-centric websites) Signs your new IA doesn't work: You need a popup announcement bar that links to an unformatted, out-of-nav landing page describing the new content. Also want to point out that I'm an equal-opportunity web-industry critic! I'm hypercritical of everyone's IA, not just SEO and media folks. I expect UX researchers to know better.

10

4 Comments

We all know these feelings: - An alert goes off. - Logs are piling up. - The team is scrambling. - No one knows what broke and why. That’s the old way of running systems. I’ve just started a new series to help teams move forward with AIOps. Here is Part 1 where I’ve broken it down covering : - What AIOps really mean? - Why it matters for every tech team? - Tools that can help you get started - How it can reduce stress without replacing anyone? So, if you're a DevOps engineer, SRE, or leading a tech team, this series might help you breathe a little easier. Read Part 1 here: https://lnkd.in/gK7xuETv Would love your thoughts and feel free to share how your team is handling ops today. Let’s build better systems together. #AIOps #DevOps  #TechLeadership #SiteReliabilityEngineering  #InfraOps #Observability  #Automation  #DigitalTransformation

2

Everyone keeps talking about writing more code like that’s the real problem. It’s not. Companies don’t need more code. Instead, they need to manage the knowledge inside their engineers’ heads. In this article for DZone I lay out a simply solved but rarely acknowledged challenge: software fails not because teams aren’t smart enough, but because organizations treat coding like output instead of knowledge. Tracking hours and pushing frameworks doesn't fix the real issue but managing the flow of understanding between humans does. If you want software that actually ships and keeps working, start thinking about knowledge transfer, learning curves, and communication before you worry about lines of code. Read more and let me know what you think and if you encounter this challenge as well: https://lnkd.in/erVZMv6R #softwareengineering #devleadership #knowledge #productivity #softwareprocess

16

3 Comments

When we talk about composable CDPs, it’s easy to focus on tech like APIs, data layers, tools. But I think the real story is human. Composable means your teams aren't locked into someone else’s roadmap. It means your analysts can work in the tools they love. It means your marketers can spin up segments without filing a ticket. It means IT isn’t playing whack-a-mole with yet another SaaS platform. It’s flexibility, but it’s also autonomy. That’s what people want in their workflows. And when you give teams the right architecture, they do better work faster, smarter, and with more impact. The future of data isn’t just composable. It’s collaborative.

15

1 Comment

What if I told you Kyber migrates templates 99.4% faster than legacy CCMs? Ya, wild. But true. We’ve seen it take 40 hours to rebuild a single template in traditional systems. In Kyber? 15 minutes. (And that’s a conservative estimate.) Most carriers have 300+ templates. So the math is… brutal. A full rebuild can take over a year, cost six figures, and still leave you chasing parity with what you already had. That’s why we built the AI Template Import Tool. It lets carriers upload their existing templates, automatically maps logic and layout, and gives business users a clean, editable version in minutes. No dev work. No additional hired help. No extra line items. Kingstone Insurance Company used it to go live in 2.5 months. Branch used it to collapse 500+ templates down to 100 using dynamic logic. The clip below shows what implementation actually looks like in Kyber. We handle the data bindings, population logic, and more. Your team just reviews, tweaks (if they want), and hits publish. Implementation shouldn’t be the hardest part. It should be where momentum starts.

15

7 Comments

Nobody talks about this, but here is what actually makes Adobe Web SDK sing on real projects - Your XDM schema? Get it airtight before you mark anything as "Profile-enabled." If you lock in fields too early, you will be patching messes downstream - trust me, rebuilding an XDM schema after mistakes is not an afternoon job. Start "base," keep it in a sandbox, break it, rebuild it, then take it live. - Ditch the copy-paste circus with mapping tables. Don't build a new rule for every click, event, or pageview. Create a single Data Element that pulls from your source. Map it with a centralized table - one lookup, not a hundred tweaks. Now your data layer is a source of truth that scales as fast as your site changes. - QA like you mean it. Seeing data in the Debugger means nothing if your XDM payload silently drops fields. Pause and check - Your data layer live in the console - does it match what is expected? - Run the payload through Experience Platform Debugger - catch those type mismatches, every time. - Wrap up with Project Griffon (aka Assurance) and make sure the XDM fields land in props/eVars, just like you planned. Follow these moves and your Web SDK is not cobbled - it is clean, fast, and future-proof. What is your "never again" lesson from Web SDK rollouts? #AdobeAnalytics #WebSDK #AlloyJS #XDM #AnalyticsAutomation

33

2 Comments

Engineering leaders don’t burn out from projects. They burn out from interruptions. - Client emails mid-task. - Calendar invites breaking your flow. - CC threads dragging you into conversations you don’t need to be in. Every interruption looks small. But stack 50 of them and your day is gone, before it even started. You don’t lose hours. You lose momentum. You restart. Again. Again. Again. And by the end of the day, you’ve been busy, but nothing actually moved forward. That’s the real cost. Not time. Execution. And the fix isn’t better focus. It’s fewer interruptions reaching you in the first place. That’s exactly what Stream does. Lemme show you how it works: https://lnkd.in/gxJ4iQWu

16

14 Comments

​Anthropic recently launched the Claude Plugin Marketplace for both their enterprise application and their developer tools. They're moving at lightning speed, and I genuinely dont know how we are going to keep up! ​Here are some of the new tools currently available in the marketplace: ​Firecrawl: This developer plugin extracts text data from websites and converts it into a clean format that Claude can read automatically. ​HR Plugin: This enterprise module automates administrative personnel operations across the entire employee lifecycle without human input. ​DocuSign Connector: This integration processes legal agreements and manages automated signature workflows directly within the Claude interface. ​Superpowers: This engineering tool provides Claude with specific system instructions to improve its accuracy on complex software development tasks. ​What This Means For Us: We're transitioning from performing individual tasks to managing automated systems. Because these plugins allow Claude to connect directly to our external software platforms, we're gonna spend less time executing manual data entry and more time reviewing the final outputs produced by the artificial intelligence. ​Please subscribe to the newsletter at https://bit.ly/RubixTech for future notifications about AI software updates, and share this post on your social media accounts!

3

2 Comments

Writing good code is one thing. Writing good pull request summaries? Who has time for *that*? But, as any delivery team knows, good documentation speeds up cycles and reduces bugs... Lately, I’ve been using this prompt inside Cursor to auto-draft PR summaries. It’s not perfect, but it's usually a good start and nudges me to think like a reviewer — and that’s 80% of the win: ``` You are a technical writer used to working with busy engineers.  I need a summary for a pull request, describing all of the changes from `main` to the current state.  Can you prepare a bullet list of notable changes? Make sure to provide a bit more context of the meaningful changes and, for some of the miscellaneous changes, just make note. Make sure, too, to describe what the tests curently cover and ensure that the tests cover any major functionality in the PR. Consider how a reviewer might do manual QA as part of the review.  Generate markdown text for easy copypasta. ``` I’d love to see how others are doing this. What’s your favorite prompt for better workflows? #shareYourPrompts

8

Documentation that isn’t maintained is technical debt with better grammar. Outdated docs mislead faster than no docs. Accuracy matters more than completeness. #Engineering #Leadership

7

1 Comment

Everyone's declaring MCP dead. I strongly disagree. Perplexity dropped it. Garry Tan called it bloated. The consensus: CLI is 10-20x better. Here's the thing. CLI wins for local coding agents. I won't argue with that. Less overhead, cleaner syntax, agents speak bash natively. But coding agents aren't the whole story. I run MCP in production. Not for coding. For business operations. Our AI connects to: • Data warehouse — real-time sales analysis across hundreds of stores • CRM — customer segmentation, promotion targeting • Push notification platform — auto-generated campaigns • E-commerce channels — synchronized listing management across 4 platforms Try doing that with CLI. You can't. Because CLI is local-first. It assumes your agent lives on a developer's machine, talking to a codebase. MCP solves a different problem: connecting AI to the systems where business actually happens. The evolution looks like this: Chat → MCP → CLI → Cloud Agents CLI is hot right now because it nails the developer workflow. Google Workspace shipped CLI. Obsidian shipped CLI. Basecamp is building CLI instead of AI features. I've seen this pattern before. I love Obsidian. Local-first, powerful, great for agents. But I still use Notion. Why? Cloud is too convenient. I don't want to set up a server to run my agents. I want them on cloud, always available, working while I sleep. CLI is a stepping stone. Cloud is the destination. The companies calling MCP dead are optimizing for token cost on coding tasks. Fair. The companies quietly deploying MCP are connecting AI to their entire business stack. Data pipelines, CRM, inventory, promotions, content distribution. That's not bloat. That's infrastructure. The question isn't "CLI or MCP?" It's "Are you building agents that only write code, or agents that run your business?" If it's the latter, MCP isn't dead. It's just getting started.

3

A business built on "vibed" code has no moat to speak of. If it's true that "A.I." enables *anybody* to build software just by describing it (it's not, but let's run with it for now), then *anyone* can reproduce your software easily. Real competitive advantage would then be found not in what these models can generate, but in what they *can't*.

43

6 Comments

🛠️ Distributable Agentic Developer Skills So, there's a bunch of these "awesome skills" repositories popping up to help nudge Agents in more useful directions, and I think distributing these is a really compelling way to have impact. Think an "instrumentation skill" or "RPC guidance skill" or similar. I think Claude has a marketplace for skills, but I certainly do not have access to that. However! 💡 Thanks to the hard work of standards writers in the past, we do have the OCI spec and a series of OCI compliant registries 💪 This includes **github**. In principle, you can store skills there, and distribute them ...fairly easily. To hook all this up, I vibe coded a quick tool — skr — to link these two together. You can check it out here: * https://lnkd.in/dkYKMBKz There's also a marketplace skill to convert your existing `skills/` repositories to a nice OCI version: * https://lnkd.in/d_hAkdDV You can even commit a `.skr.yaml` file to your project, so you can share skills with your peers! This helps you: 1. 🚚 Distribute best practices _as technical guidance_, right in the app. 2. 📢 Share instantly your lessons (e.g. postmortem) with your peers across the group. 3. 🚀 Build knowledge of your own technical platforms right into the vibe coded hands of your engineers. It's also nicely enterprise friendly as everyone has an OCI compliant registry ... somewhere. It should just instantly work! Check it out on my (very limited) skills repo here: * https://lnkd.in/dVnezdPE Let me know what you think ♥ it's only about 4 hours of work. I'm very confident there will be bugs, but it's pretty cool how fast this came up.

13

2 Comments

I don't mean to generalize. But if you're app is less than 50,000 lines of code I don't understand how you continue. With ever-more over arching Claude (and others) contracts being signed, it seems to me those 10,000 lines of code can be replaced in 2 steps, a decent prompt and 2 iterations.

30

108 Comments

After years working in API-first universes, and now that we live in the era of MCPs, I can't get my head around the fact that availability / booking systems of hotels "only" work on GDSs (Amadeus, Sabre, etc.) or proprietary extranets (Four Seasons Preferred, MO Fan Club, etc.) and don't offer a kind of unified API layer. Or am I missing something?

9

User training is where most technology implementations fail. You buy great software. Your team gets 2 hours of training. Then they're expected to be productive immediately. Reality: they use 20% of the features and work around the rest. This isn't a training problem. It's a design problem. Software should be intuitive for people who actually do the work. Environmental health inspectors shouldn't need to become database experts. PH360's interface follows public health workflows, not software logic: → Inspection processes match real-world procedures → Forms are organized like paper versions teams know → Guidance is contextual and job-specific → Advanced features are available but not required Make technology adoption natural, not forced. Because the best software is the software your team actually wants to use. How long does it typically take your staff to become proficient with new systems? -- Hi! I'm Jefferson the CEO of F&T Labs. I'm a former CIO of a local health department and I've taken that experience and funneled into the tools that we're building at F&T Labs. I envision a future where America's public health systems have the tools that they need to do their work. (I'm also a proud husband, dad, and bow-tie aficionado)