Elastic's Chief Product Officer Ken Exner on AI and Context Layer

The most important promotion of your career just happened... and you didn't even apply for it. Ken Exner (CPO at Elastic) just redefined the developer career path in under 60 seconds during his conversation with our founder, Bernard Leong in the upcoming episode of Analyse Podcast. He argues that the future of software development isn't about being an Individual Contributor (IC). It’s about becoming a Manager of Agents. Think about that. Your "team" isn't just the person in the next cubicle—it’s a suite of autonomous agents. Your new "coding" language? Context Engineering. The New Dev Workflow: 1️⃣ Context: Passing the right data/environment to the LLM. 2️⃣ Intent: Setting clear, high-level goals. 3️⃣ Audit: Reviewing AI-generated output for accuracy and security. This is a massive shift in mindset. We are moving from "doing" to "directing." Full episode drops tomorrow here: https://lnkd.in/gqK8MvEg Question for the devs out there: Does this excite you or terrify you? Let’s discuss in the comments. 👇

Just finished recording a nearly two-hour podcast as a guest on What The Tech (AU) with Vivienne and René. A few of the views I shared: • Every AI app is a data app in disguise. • Developers do not want less control. They want less plumbing. • Most engineering teams do not have a scaling problem. Most have a complexity problem. • AI lets more people build. Solid foundations are as important as ever. • Open source is becoming an almost existential choice for enterprises. • The market is moving away from distributed by default and back towards simple stacks that earn complexity later. That is why people now launch businesses in 36 hours, not months. • Building distributed applications before you have millions of users is often better for your resume than for your product. Stay tuned!

Most companies don't have a data problem. They have a communication problem disguised as a data problem. I just sat down with Sean Sullivan on the Superhuman Podcast - a marketing and revenue operations leader who's spent over a decade helping companies scale without the chaos. Here's what hit hardest: He told a university professor teaching Python and regression models that most companies actually need SQL first - because their data is a mess before they can predict anything with it. We got into why teams stay misaligned even when the tools are there, the difference between "pray and spray" vs. "land and expand," and why the best leaders don't pull people along - they remove barriers. Sean also dropped a line I won't forget: "You're a human being, not a human doing." If you're a founder or operator feeling buried by noise instead of guided by clarity - this one's for you. Full episode link in the comments. Stay Superhuman. –– Each Saturday, I share one idea to help you feel better and operate at your 1%. Join 5,000+ readers → https://lnkd.in/eQ-4nEhg

Wes McKinney's dropped this thesis in a Data Renegades Podcast: AI has created radical accountability for every software vendor. Building software just got dramatically cheaper. One engineer with a Claude subscription can prototype a replacement for tools that entire teams used to tolerate. Customers no longer have to accept mediocre products because the cost of leaving has collapsed. Wes's message to vendors shipping broken tools: "This is bad. Why haven't you fixed this yet? If I was on your engineering team, I would have already fixed this. I would have done it like today with Claude code." The flip side: a flood of AI-generated projects that only make sense to their creators. Hyper-personalized software with bad taste. The barrier to entry dropped, but the bar for credibility rose. "People are going to decide which companies to pay attention to on the basis of how credible the people are involved." Full episode linked in comments, or wherever you listen to podcasts #AIcodingagent #radicalaccountability #agenticworkflows #dataengineering #DataRenegades

Genuine connection is your greatest asset and it can’t be automated. After over 25 years as a software developer, scaling teams and solving complex technical problems, Michael discovered the greatest asset was simply human. Michael Blood is the founder and President of Matraex, Inc, a software development company focused on turning ideas into working systems. We discuss the importance of passion, breaking things, and how humanity is the advantage in an AI world. A few key takeaways from our conversation - Start doing things and quit what you don’t like, fast - Be willing to break stuff - Authenticity can’t be automated - If you’re not humbly learning, you’re losing - Your biggest blindspot and largest asset is standing right beside you The Matthew Byrd Podcast Listen to the full conversation below👇

One of the big topic areas for Risk-First Software Development is AI. How does risk management apply to a potentially world-changing new technology? In this excerpt from the JUXTCast Podcast, Henry Garner and I break it down. Check out JUXTCast here: https://lnkd.in/e9CjMJrD Get hold of a copy of the Risk-First Software Development book! https://lnkd.in/eVck5xgy

Just finished the 3+ hour deep dive with 🦄 Peter Steinberger (OpenClaw) on the Lex Fridman Podcast. Besides the passion he works with, I extracted from him a few practical tips on working with AI agents 👇 1️⃣ Treat it like a conversation Don’t just bark "do this and that" commands. You can actually use it as a sparring partner to think through things. 2️⃣ Tell it to slow down “Take your time.” I was surprised by this one. 🤷♀️ Peter explains that apparently the AI rushes through answers when it freaks out that it is closer to its context windown. 3️⃣ Point it where to look The agent doesn't keep the logical overview of your docs and repo. You get better results by directing it where to find the info: “You need to consider A, B, and C." I've hit some walls with that one. 🤪 4️⃣ Think from its POV Again it doesn't have the human overview of how things relate to one another and its context limits are real. Peter was talking about understanding how the tech works so you can be more "understanding/emphatic" when directing it. 5️⃣ Force self-introspection I've had problems with it saying it did things it didn't and it was simply because it wasn't "seeing" the page I was looking at. So Peter mentions how helpful it is to ask it: “What tools do you see?” “What error are you getting?” “Read the source. What’s actually wrong?” And then make it debug itself. Not assuming that it sees the same errors as you do. 6️⃣ If it’s stuck, stop If it ends up in long execution time likely something is going wrong. Either you gave it weak context, or your architecture is messy. So re-think the approach. 7️⃣ Let it write its own AGENTS.md I really liked this one. ❤️ Tell it how you want to collaborate. Then say: “Write your own AGENTS.md. Give yourself a name.” Peter thinks that the ownership you infuse in the agent can increase its quality. 8️⃣ Make it self-aware It should know: Its own source code, the docs, the model it runs on. He says this helps it perform better and get less lost. 9️⃣ Review only what matters This one we've stumbled upon a lot lately. With an agent that produces 174 PRs overnight, you have to spend your entire morning reviewing them because PRs and approval processes are still made for human-to-human collaboration. Peter suggests to approve by default anything that is not: - DB queries - Data mutations - Anything irreversible 🔟 Add heartbeats Peter literally started with: “Surprise me.” Every 30 minutes. To me this proactivity of the agent was the magic that sold it! 1️⃣ 1️⃣ Skills > MCPs MCPs bloat context. CLIs are natural. Models already understand Unix patterns. A skill = one sentence description. Details load only when needed. ... Do you have other small tips that make agents work better for you? Let's get the collective wisdom going. 👍 👍 #OpenClaw #AIAgents #AgentBestPractices

Coffee Thoughts ☕️ Three questions can dramatically improve GenAI output One of the easiest ways to get better results from GenAI is to stop asking it to read your mind. A lot of weak output happens because the model starts generating before the problem is actually well-formed. The ask is too broad, the audience is unclear, the real objective is fuzzy, or there are hidden constraints sitting in your head that never made it into the chat. That’s why I keep coming back to a simple move that has outsized impact. Before GenAI does the work, have it ask you three questions. Usually those questions get at things like: • who this is for • what success actually looks like • what constraints, tone, or tradeoffs matter most It sounds small, but it changes a lot. The process becomes less “generate something and hope” and more like a quick Socratic setup where the model helps you sharpen your own thinking before it starts producing. Better inputs, better framing, better output. It also helps in a more subtle way. Those three questions often surface decisions you hadn’t made yet. Do you want one answer or three options? Do you want speed or polish? Is this for an executive, a customer, or your own scratchpad? Clarifying that up front saves a ton of downstream rework. I had a fantastic discussion with Joshua McCay, MBA the other day on strAIght talk. I’ll drop the episode link in the comments if you want the fuller breakdown.

Applying for Tech roles these days feels like signing up for a Netflix series you didn’t know had 9 episodes 😄 Episode 1: You send your resume into the void. Episode 2: An AI reads it in 0.3 seconds and says, “hmm… interesting.” Episode 3: Recruiter call — you try to explain your entire career like a movie trailer. Episode 4: Take-home assignment — suddenly it’s a full weekend project. Episode 5: LeetCode round — “reverse this binary tree while solving world peace.” Episode 6: System design — you casually design something used by millions. Episode 7: Culture fit — “tell me about a time you were a good human.” Episode 8: Wait 3 weeks. Character development arc. Episode 9: Plot twist — rejection with no feedback. Season ends. You stare at the screen like… “same time next week?” 😂 But honestly, if you’re going through this: It’s not just you. And the funny thing is — some of the best engineers I know have gone through multiple seasons of this show. So if you’re feeling a little low during the process: Take a step back and remember — you’re learning, growing, and getting better with every round (even the weird ones). Your story isn’t defined by one application… or one season. Season finale is coming. And it’s going to be worth it 🚀 #AI #MachineLearning #DataScience #SoftwareEngineering #JobSearch #TechHumor #Careers

One of the things that excites me most is programming creatively. I built a podcast tailored to my interests — no studio, no mic, no editing. I built it using: 📰 Forbes Brasil scraping ⚡ Aspire .NET 🎯 Blazor 🤖 Ollama + Gemma 3 🎙️ Kokoro TTS A bit of creativity and some open-source tools, and you can have a personalized podcast (pt/en). Tech is so accessible today, thanks to the people who share their creative code with the community. What would your AI podcast be about? 👉 Check out the full project here: https://lnkd.in/dgGvi9Gi