Jeff Auston

Agents are the new apps. And they just went multiplayer. The software engineering landscape is undergoing a seismic shift. Google just dropped Firebase Studio. And it might just replace half the AI builder startups out there. You’ve probably seen tools like Lovable, bolt.new, Vercel's v0, or Cursor. They let you build AI-powered apps without writing much code. Describe what you want — boom — they create UIs, workflows, and automation for you. It’s fast. It’s smart. It’s changing how software is built. Now imagine that… backed by Google. What is Firebase Studio? Firebase Studio is Google’s new end-to-end app builder for AI-first products. It combines: ✅ UI generation ✅ Agent workflows (like AgentSpace) ✅ Firebase backend (auth, storage, hosting, etc.) ✅ Google’s best Gemini models AgentSpace is like a command center for AI agents. It lets you create multistep workflows, connect APIs, and manage how agents behave. Great for product teams prototyping internal tools, automations, or AI assistants. But it’s still pretty early-stage. Right now, Firebase Studio and A2A (Agent-to-Agent) are two separate moves from Google — but they are strategically aligned, and integration is very likely coming. You need to manage agents, instructions, tools, and sometimes code. Why this matters for product managers? → The barrier to building software is decreasing. → With Firebase Studio, a product manager can do what used to take a team. The AI agent is becoming the new app. Not just chat. Agents that do stuff: send emails, fetch data, run logic. Google wants to standardise agents across platforms. Their new Agent-to-Agent protocol (A2A) means your agents could soon talk to each other — across apps, tools, maybe even companies. Enter MCP: The USB-C for AI Anthropic’s Model Context Protocol (MCP) is an open standard that enables developers to build secure, two-way connections between their data sources and AI-powered tools. MCP is a standardised protocol designed to streamline the way AI models communicate with external systems, eliminating the need for custom integrations. With so many advancements happening rapidly, it’s challenging to keep up and navigate them within the realities of development. Yet, it has never been more exciting to be in the software engineering space. #AI #FirebaseStudio #ProductManagement #Innovation #TechTrends #MCP #A2A

16

1 Comment

Adding Document Understanding to Claude Code 📑🤖 Coding agents are awesome but they don’t natively understand docs (they kind of do, but it sucks). This limits their ability to build business applications - think about anything from a financial copilot to contract reviewer to rfp synthesizer to synthesizing PRDs. I wrote a blog to capture core patterns for adding document understanding to Claude Code - applicable to coding agents in general. This both helps Claude Code 1) get enterprise context when building apps, and 2) allows it to build more agentic, robust apps in general. 1️⃣ Add an MCP server that preindexes your file knowledge base 2️⃣ Let Claude Code loose on your filesystem with the right CLI tools for understanding files. 3️⃣ Give Claude Code context on core modules for document understanding and agentic workflows - this last step is important if you want to build “better apps in general” and not just “rule-based apps”. Blog: https://lnkd.in/g4EEFftk The blog offers general principles regardless of whether you’re using our stuff, because we fundamentally think that coding agents + docs leads to a ton of useful applications. Of course we are deeply investing in tooling in this space too (some referenced in the blog), a lot of it dependent on LlamaCloud. Check it out: https://lnkd.in/g9Wpqn7w

213

2 Comments

I had a heap of great questions to yesterday's context engineering post so here's a followup with more detail on how we think about core responsibilities / ownership ** Product / Context Architect → designs the end-to-end context layer, balances accuracy vs. cost, and sets KPIs for “return on context.” This person might be your main owner as they can see the whole picture of your AI system ** Domain Subject Market Experts → curate and continuously validate reference knowledge so retrieval never serves stale data. These folks are often the secret glue you can use to ensure you have the best data and it stays accurate with the world as it moves ** ML / Data Engineers → build the chunk → embed → retrieve → re-rank pipeline and tune it for latency + token ceilings. Obviously, the golden bandana of all AI / ML integrations ensuring fast pipelines / access to tools / model performance ** Prompt / Interaction Designers → craft system & task prompts, tool-call patterns, and fallback messages that keep UX coherent. Each of these folks are product people in their own right ensuring the agents, prompts, and overarching communication strategy is strong ** Security & Compliance → tag sensitive snippets, enforce PII redaction, run jailbreak drills, and maintain audit logs. Often you should be able to tag your overarching business compliance and security into the context engineering workflow but it's good to have a person who is thinking about these elements as it relates to each involved AI system ** SRE / MLOps → watch inference latency, token spend, and infra scaling; own emergency rollbacks when a context tweak nukes prod. This is a magic trick for efficiency if your MLOps teams can also be thinking about token spend / latency as you can save cost on tasks that can take longer to run with sight into internal / external models ** QA / Eval Engineers → automate evals (faithfulness, toxicity, bias) and gate releases; feed failure cases back into retrieval or prompt fixes. This team should be super tight with the Prompt / Interaction designers and ultimately also regularly share how well the system is at hitting 'must haves' vs 'nice to haves' vs 'sure why nots' ** Data-Governance Lead → owns lifecycle rules such as versioning, retention, provenance for every knowledge chunk. Definitely similar camp to security and compliance but closer to the specific data being fed into this instance of the model. They should be governed by the overarching data strategy but hyper aware of what's in this specific system. ** Customer-Success Liaison → funnels real-world hallucinations or policy breaches back to engineering so they actually get fixed. This team is in the users pocket further than any other person in the stack, they live and breathe customer feedback and obsess over edge cases. ** What else do you tend to sit under context engineering?

220

5 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

🔒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

241

3 Comments

How Vanta Uses AI w/ VP of Engineering Iccha Sethi! 10 min deep dive. We cover: - Hiring engineers for the AI era - AI unit tests for large code bases - Writing docs with AI - Vibe coding for eng leaders - Favorite AI tools

134

4 Comments

💡 AI doesn’t just change how code is written – it changes expectations and outcomes. In the final section of Jellyfish’s AI Adoption Guide, we explore how expectations for engineering leaders are evolving as AI becomes embedded across the SDLC. Because the organizations pulling ahead are treating AI as a transformation, not a tooling upgrade. Inside, you’ll explore: - Why AI adoption is quickly becoming a leadership mandate, not an option - How to balance ambitious goals with realistic expectations - What cultural investments are required to sustain long-term gains - How data can be used to align executives and engineering teams The teams that win in the AI era won’t just have better tools – they’ll have leaders who know how to guide change. Download today: https://lnkd.in/dut_6WpQ

19

Steve Yegge posted a blog in march that encapsulates the massive shift occurring in software engineering today. With Claude 4 and the recent improvements on Claude Code, we're entering a new phase. https://lnkd.in/gRtRidBr This year, I have pushed more code than you. But, ideally, I wouldn't need to. We are now in the phase of AI software development called the "Agent Clusters". Here's how it works: Each time our code in production fails, a github actions claude code job auto-starts to fix it. Each time we want a new functionality, as we're communicating over slack, I can tag in our slack bot to start a github actions claude code job. And I can make as many run in parallel as I want. The bottleneck has shifted to how many things can you think of to code at a time, and how to write a good meta-prompt like a CLAUDE.md or good cursor rules files to have good code architecture over many AI iters.

60

20 Comments

Here's the 80/20 Setup & Understanding to use Claude Code for GTM Foundational concepts: Claude code is a CLI tool - that means command line interface. That means "looks intimidating empty black screen with colorful words but is something you are 100% capable of figuring out and the logic to use it will quickly feel familiar once you get the hang of it" Installation: Anthropic recently released a one line install (check the link but literally just copy/paste based the system you use). I still want to break down the old way to install to help you understand software better 1. Install Node.js (Javascript for the server) This means like 15 years ago someone wanted be able to run Javascript outside of their web browser (it was made just for web browsers), so they created a whole way to run Javascript for servers and (importantly here) inside a command line tool. 2. An IDE. This means Integrated Development Environment. What this really means is Cursor. Cursor is just an existing IDE (VSCode) + AI. VSCode was a go to IDE for most devs before the AI boom (myself included), now everyone uses Cursor. Once you install node (its like installing any other app on your computer with the installer app) You first open cursor Then you type this: ``` npm install -g @anthropic-ai/claude-code ``` Lets break this down: npm = Node Package Manager NPM comes with Node and is basically does what the app store does for your iphone. "-g" says install this globally (so not just one project, on your whole computer) and that last bit is like your mailing address, it tells npm what to get and what to install Why do you need Cursor? At base, all an IDE does is open a specific folder and show you everything in there, and make sure your terminal is talking to the same folder. This is really useful for developing apps, since you've got everything sync'd up. To make sure Claude Code can read and write to a central spot, you open it inside Cursor so you can keep track of everything Ok so once you did that (or just ran that command), to do anything in Claude Code you should know three main things. 1. You start it by typing "claude" in your terminal 2. you do stuff with a slash " / ". This opens a list of commands you can select from. 3. Shift + Tab changes the mode of Claude code (there is plan, auto-accept edits, and manually approve edits) 4. Double tap Esc to go back and edit a message (same way in chatgpt/claude you can click the edit button to edit a message you sent) 5. Finally ctrl+c to close Claude Code. (actually 5 things I lied oops) That's the basics to set it up. If there's interest in more I'll put out more advanced content just wanted to get a vibe check for where ppl are at with it 🙂

25

6 Comments

Engineering Management in 2025 is 𝘊𝘩𝘢𝘯𝘨𝘪𝘯𝘨. The end of ZIRP (Zero Interest Rate Policy) and the rise of generative AI have sent ripples across the industry. Leadership assumptions that were true last year may be wrong next year. You read posts on how this impacts software engineers. But what about software engineering managers? ❓ What is changing? ❓ How can you use LLMs for your job? ❓ What should you avoid using LLMs for? ❓ What will stay the same, even a decade from now? Check out: "How to Manage in the Age of AI" April 2nd, 12 PM EDT sign-up for free: https://lnkd.in/exFsAegW

32

18 Comments

What is agentic delivery? Imagine a world where every software implementation starts aligned— and stays that way. Where discovery isn’t buried in notes, where scope changes don’t break everything, where tribal knowledge isn’t lost between teams. Imagine if: 🧠 Requirements were structured as they were spoken 📄 Artifacts were generated in real-time 🔁 Context followed every decision, across every tool ⚙️ Change was handled with traceability, not fire drills That’s what agentic delivery unlocks. It’s not just AI stitched onto old workflows. It’s a new operating system for how delivery gets done— powered by agents that structure, generate, and evolve work with you. Not to replace people. But to give your best people the leverage they’ve never had before. → Fewer handoffs. → Less rework. → Stronger knowledge reuse. → Faster, more confident go-lives. This is the future of services. And it’s already being built. Welcome to the era of agentic delivery.

34

4 Comments

🚀 Announcing timveroOS 8.0 - Deploy Your Customized Lending Platform Faster The traditional lending platform deployment timeline just got disrupted. Version 8.0 introduces architecture that takes you from decision to working system in as fast as three business days—fully customized for your credit products and workflows. Key highlights: ✅ Rapid Deployment - Working customized system in 3 business days, not 6 weeks ✅ Payment Hub - Automatic allocation of consolidated payments across multiple loans ✅ Complete Audit Trails - History tracking for regulatory compliance and transparency Financial institutions evaluating lending technology can now pilot faster, launch quicker, and scale without infrastructure complexity. Operations teams get intelligent payment processing that turns 20-minute manual allocations into 2-minute automated workflows. Read the full release notes: https://lnkd.in/dshAAUGd

24

1 Comment