Ryan Marshall

This is a really interesting article about writing agents with the Claude Agent SDK. It gets into design advice that you should consider when writing an agent with Claude. It chimes with what I’ve read about the development of Claude Code (write as little business logic as possible and concentrate on getting out of the way of the model). The general takeaway for me is from the three-block flow chart in the middle: Give the agent good tools/affordances in three categories: Gather Context, Take Action, Verify work. Do that at the right granularity, with good feedback, and then let the model do the driving. It’s deceptively simple, but it has a goldilocks quality to it. Trying to make a more complex flow chart with many more boxes and lines constrains the AI to just the things you could think of in advance and makes it overfit to what you know today. Trying to make it simpler by, for example, giving the agent a whole existing API as a single tool can overwhelm it with too much choice and then too much context in the responses. It’s definitely worth reading if you're interested in building or using agents, irrespective of AI vendor. https://lnkd.in/eKF2rXWb

7

I've been using Claude Code plugins for months. Code review, client project management, building shared context that compounds over time. Less a tool, more an extended cognition ecosystem. A few days ago, Anthropic brought that same plugin architecture to Claude Cowork, and it's worth paying attention to. Here's why. Claude Code let developers shape AI to their specific context, workflows, and accumulated knowledge. Plugins made that shaping persistent. You teach Claude your world once and it shows up informed every time. But it required a terminal. Cowork plugins remove that barrier. Same composability, same feedback loops, now accessible without writing a line of code. 👉 This is the shift that matters: thinking with plugins and feedback loops, not static prompts. A prompt is stateless. When you switch chats, context evaporates. A plugin encodes your processes, your terminology, your decision frameworks. It persists. It compounds. And that compounding is what separates AI-native organizations from everyone else. https://lnkd.in/gJZEtNpR

15

1 Comment

So which models produce HTML that has the least accessibility errors, and at what cost? Mapped it out here (bottom left is the best). This is helpful for understanding which models are most compliant for those with disabilities and do it efficiently. Qwen and Gemini 2.5 Flash lead the pack. Hat tip to Ben Ogilvie for pointing me to this: aimac.ai

55

2 Comments

Innovation often hides behind wrappers. MCPs promise scale and speed, but let’s be honest: they’re just orchestrated prompts and APIs. Valuable? Yes. Foundational? No. The deeper question: are we funding primitives—or just better packaging? The history of technology is littered with companies that confused distribution hacks for paradigm shifts. Efficiency matters. But leaders must know the difference between an infrastructure play and a true leap in capability. So—what deserves our capital: elegant wrappers or new primitives? https://lnkd.in/g8yUQzDb #AI #InnovationStrategy #Technology

14

2 Comments

I thought this was a good and very reasonable talk by Eleanor Millman on how to prioritize projects for a platform engineering team: https://lnkd.in/gaYtUHRH A few insights I took from it: - Unlike product features which can be more directly tied to revenue, platform engineering work is a couple steps removed. But that doesn't mean it's unimportant! Far from it, platform engineering projects can make everyone at the company more efficient, able to produce higher quality software, etc. - While "urgency" is always going to be a factor, you don't want to just be putting out fires, you want be able to prioritize work which is "high impact" ... even if that impact isn't felt immediately. - Good rule of thumb: prioritize "highest impact for lowest effort". - The talk has some ideas on what factors to choose for impact, and how to blend them. For instance "speed of development" is one factor, "cloud cost optimization" might be another. The weighting of different impact factors can change over time, depending on the needs of the business. I would say that a lot of companies don't have much methodology here but I can really see the value in codifying it. You can always change the methodology or the weighting if it's spitting out results that don't pass the smell test or you really feel it is leading the company astray. Clarity can give the org more freedom to put resources behind projects that would otherwise never be taken on. When things are unclear, prioritization still happens implicitly, but the decisions tend to be a lot more random and fear-based and the org is worse off as a result.

15

3 Comments

ICYMI: v0.dev is now v0.app and that means a lot more than you think. Vercel just made a smart move that many people haven’t picked up on yet. Sure, v0 is better at generating code. And yes, it’s more “agentic” now. But if that’s your takeaway, I think you’re missing the real move. The shift from v0.dev to v0.app isn’t a rebrand. It’s Vercel positioning itself to own the enterprise software creation market. v0.dev said: “This is for developers” and you built front-end components. v0.app says: “This is for everyone” and you're cranking out a CRM, a dashboard, a workflow, auth, data persistence, whatevs ... get it all built and deployed nbd. 🤷‍♂️ 💡 That’s a leap from a ~30 million developer market to a 300+ million knowledge worker market. From competing with Cursor and Copilot to taking on Microsoft Power Platform and Salesforce Lightning. 💡 This is no longer about code generation and developer productivity boosts. For coding this isn't not about "how" anymore it's about "who." #Vercel #v0 #NextJS #FullStackDevelopment #DeveloperExperience #EnterpriseSoftware #ComposableArchitecture #NoCode #LowCode #ProductManagement #ReleaseMode #FullStackDrip

60

7 Comments

I built a meeting companion/agent 3 times before getting it right. First in TypeScript. Then Electron (half a gigabyte, way too bloated). Finally Rust. It's an AI meeting companion that actually helps during calls. Hit Control + Space to start recording, get real-time transcription, and when you're done you can polish the text or generate full meeting notes with action points. I wanted something that felt like a real contributor, not just a passive recorder. Today I'm open sourcing it. Here's the full story: https://lnkd.in/ezTZC2s2

24

2 Comments

🔒How to Build an Agent with Auth and Payments Fullstack agent application! Not only can LangGraph + LangGraph Platform help you build agents, it can also help you deploy them and supports custom auth and middleware for usage tracking Video: https://lnkd.in/ghu5cGKa

240

3 Comments

In today’s Substack post from the Plan of Record series, I walk through the operational frameworks I’ve relied on to build high-performance product development teams and why they’re essential for diagnosing and fixing what’s broken. I also explain why we had to create the PoR Prioritization Framework to address a critical gap no other model could fill. Here’s the link: https://lnkd.in/ek9J_cqq #SaaS #ProductManagement #EngineeringLeadership #R&D #Prioritization #Execution #CPO #CTO #CPTO

14

I’ve just published the next installment in my "Theory of Durable Execution" series. Across frameworks we hear terms like workflows, activities, virtual objects, handlers, and functions. Underneath the terminology, they’re all describing a small number of execution patterns. In this new post, I break durable functions down into three forms: 🔹 Stateless functions for one-shot deterministic logic 🔹 Sessions for long-running, interactive orchestration 🔹 Actors for persistent, message-driven state To get there, I walk through: 🔹 The behavior–state continuum 🔹 How identity, lifetime, communication, and concurrency shape execution semantics 🔹 How frameworks like Temporal, Restate, DBOS, and Resonate map onto these forms This builds on the previous two posts in the series: 🔹 Demystifying Determinism in Durable Execution 🔹 The Durable Function Tree (part 1 and 2) This was the last post in this series, distilling the various frameworks down to a simpler logical model. Full post: The Three Durable Function Forms https://lnkd.in/eJHpY9qC

77

7 Comments

Running a Company on Markdown Files CompanyOS: a skills-only system that turns Claude Code into the operating layer for an entire company. No application code, no web UI - just markdown files that teach Claude Code how to run business operations. https://lnkd.in/gACFhTMk #AdventuresInClaude #ClaudeCode #CompanyOS

268

52 Comments

🚨 ~$1B raised in Dev + Infra across Nov–Dec. The most unhinged seed rounds I’ve seen lately. 🚨   Seed is the new Series C guess.   Almost half a billion went into infra alone - someone finally opened the AWS bill.   AGI is taking a breather, small language models are everywhere. I guess size doesn't matter after all!   Here are some of the coolest rounds:   🔥 $475M Seed - Unconventional AI Yes, seed. Backed by a16z, Sequoia, Lightspeed, Databricks, Lux. This isn’t a startup - it’s a full-on rethink of the computing substrate for AI (hardware + software + energy).   🔥 $75M Seed - Tenzai Battery, Lux, Greylock. Builds an AI-powered penetration testing platform that continuously simulates real-world attacks against your systems. Designed to proactively surface vulnerabilities as software and AI-generated code rapidly expand the attack surface.   🔥 $70M Seed - Gradium Audio-native foundation models. Develops a unified audio architecture covering speech generation, transcription, voice transformation, and dialogue, optimized for real-time, low-latency interactions rather than text-first pipelines.   🔥 $50M Seed - Inception Diffusion-based language models with lower GPU requirements. Replaces traditional autoregressive generation with diffusion techniques to improve efficiency, controllability, and latency, targeting production use cases where compute cost is a bottleneck.   🔥 $41M Seed - Mirelo AI AI-generated sound for video. Provides tools and APIs that automatically generate synchronized sound effects and music from video content, enabling faster and more scalable post-production workflows. A lot of cool stuff being built in Europe!! Shout out to orq.ai Yasu Specific (YC F25) Albatross AI INTELLIGENT CORE NobodyWho Open Machine

71

4 Comments

If you're looking for a new docs framework for your dev tool, I can't recommend Docusaurus enough. 1. You can change the appearance of any component—no matter how big or small—by ejecting its raw React code into your source tree. I've ejected almost all of its major components and restyled them with Tailwind. 2. It supports content written with MDX, which is like Markdown but better. MDX lets you write JSX tags so you can render React components (like badges and banners) directly inline your content. No more messy templating. 3. It bundles Prism for code snippet highlighting. So you if you want to change how each and every token of your language gets styled—which I do—you can override it with a simple centralized map—which I have. 4. You can swap out its on-site search engine with Algolia, now widely considered the best option for developer sites. It is, in other words, the perfect tool for people like me who are obsessed with every pixel.

118

16 Comments

While everyone was talking about Opus 4.6, for me the true killer feature of the recent Claude Code updates is the agent teams. It's been something i've been trying to achieve with customization for a while. Custom agents, orchestration scripts, specific CLAUDE.md instructions to coordinate work... with some degree of success. But what Anthropic shipped natively is a WHOLE different level. What makes this stand out is the inter-agent communication. We're not talking about simple fan-out/fan-in where you spawn workers and collect results. These agents talk to each other. Peer-to-peer messaging, dependency-aware task graphs that auto-unblock, agents that self-claim work from a shared task list. The lead can even enter Delegate Mode where it does ZERO implementation, only coordination. The image below is from one of my setups. A team manager orchestrating a librarian agent, a PhD lead, and 5 research sub-tasks with blocking dependencies. The librarian unblocks the research tasks, the PhD lead aggregates everything. All coordinated autonomously. And with this a whole new world of orchestration just unveiled itself. Distributing work across agents is the "easy" part. You break down tasks, assign owners, define dependencies. The HARD part, and what MOST stands out now, is aggregation. How do you take the output of 5 parallel agents, each with their own context window, and synthesize it into something coherent? That's the new skill. Anthropic themselves used 16 parallel agents to build a 100,000 line Rust C compiler that compiles the Linux 6.9 kernel. No human actively coding. ~$20,000 in API costs over ~2,000 sessions. We went from pair-programming with AI to managing AI engineering teams. The skills that transfer are the ones from engineering management: task decomposition, context management, knowing when to intervene vs let the team self-organize. This is a new paradigm, and i think it opens up several possibilities we haven't fully explored yet. ref.: https://lnkd.in/dVCe344z

68

12 Comments

Attending RailsConf and talking to folks, my biggest realization is just how expansive Ruby on Rails’ usage is, just that it doesn’t get spoken about publicly as often as other frameworks / technologies. Rails just seems to be silently chugging along in many companies’ tech stacks for more than a decade while happily helping the company scale. If there was one piece of tech I wish was actually hyped up way more than it is, it would be Rails, for how much productivity gains it brings to the table.

43