Jesse Talavera

Coding agents are shipping to production. But how do you know they’re doing the right thing? Coding Agent Evals 101 is a practical, engineering-led session on how teams evaluate coding agents in the real world. We’ll cover: - why evals matter for agentic systems - how coding agent evals are designed - what “good” looks like in production Maksim Shaposhnikov, Research Engineer at Tessl, will share how we design and run evals for coding agents, and how evals help teams build confidence in agents. 📍 London 🗓 Tues 17 March, 6–9pm 🔗 Link in comments

14

1 Comment

Zed is quietly shaping up to be one of the most interesting code editors for 2026. Why does it feel faster than Cursor, and could real-time Channels change how you code? Check it out!

2

🤔 Where would you see yourself if you didn't choose a career in coding? 🦪Diana Pham asked developers at Commit Your Code Conference this question, and got some surprising answers 🧀 What would be your choice?

47

12 Comments

It’s a strange time to be in software development. Not bad. Just… different. Everywhere we look, AI-built apps are taking the spotlight. Entire projects are being spun up from a handful of prompts. Job descriptions now ask for “AI tooling experience” before mentioning core programming languages. And it raises the question: Do developers even need to code anymore? Prompting is becoming the new programming. We’ve seen it used to scaffold codebases, generate tests, and refactor legacy systems. It’s powerful, fast, and surprisingly fun. But here’s the reality: the more we prompt, the less we flex the problem-solving muscles that make great developers. The landscape has shifted. Less “crafting software,” more “assembling AI outputs.” Efficient? Absolutely. Creative? Not always. Here’s what we believe: 👉 Prompting is a shortcut. Coding is still the superpower. 👉 Without debugging, testing, and understanding the why, progress stalls. 👉 Ignoring AI is just as risky; you’ll fall behind. The sweet spot lies in balance: using AI as a tool, not a crutch. The best software still comes from human creativity, not just machine outputs. Now we’re curious: How is your team approaching this shift? Do you rely more on AI to generate code, or do you still prefer to build it by hand to keep the spark alive? #SoftwareDevelopment #Coding #Programming #AppDevelopment #Developers #AIinTech #ArtificialIntelligence #TechTrends #FutureOfTech #Innovation #DevCommunity #TechLife #CodingCommunity #BuildInPublic #DigitalTransformation

2

Ever hired a “10/10 resume”… and still ended up debugging production yourself on Monday? 😅 In emerging tech, resumes age faster than software versions. Buzzwords look impressive. Real problem-solving? That shows up only when systems break. The truth is — Great engineers aren’t defined by how many tools they list, but by how they think, adapt, and build under ambiguity. Especially in data engineering, AI, and cybersecurity, the gap between knowing and doing is where most hiring mistakes hide. Smart teams are quietly shifting from resume-first hiring to capability-led hiring — focusing on system thinking, real-world execution, and learning velocity. That’s how scalable tech teams are actually built. At The Briminc Softech, this is exactly where we come in. We help TA teams and senior leaders cut through resume noise and connect with engineers who can build, secure, and scale — across data engineering, emerging technologies, and cybersecurity. Because when the right engineer joins your team, systems ship faster, outages reduce, and hiring finally feels… logical. 🚀 #TechHiring #ITStaffing #TalentAcquisition #EngineeringHiring #TechLeadership #HiringStrategy #FutureOfWork #TechTalent #SmartHiring #LeadershipInsights #TheBrimincSoftech #StaffingSolutions #BuildBetterTeams

9

A few months into my first dev job, a senior engineer on my team said something I've thought about ever since. "Junior developers write code that works. Mid-level developers write code that works and is understandable. Senior developers write code that works, is understandable, and is ready for what comes next." At the time I was focused entirely on the first part. Making it work. What I've learned since then is that the jump from junior to mid-level is almost entirely about the second and third parts. The second part — understandable code — is about naming, structure, and documentation. Variable names that describe what they hold. Functions that do one thing. Comments that explain why, not what. TypeScript types that make the shape of your data obvious without having to read the implementation. The third part — ready for what comes next — is about architecture. Writing a function that's easy to change when requirements change. Structuring a database so adding a new feature doesn't require rewriting the schema. Building an API so the frontend can evolve independently. These are not advanced concepts. They are habits. They develop by being intentional about them from the beginning. The next time you write a variable, ask: does the name explain what this holds? The next time you write a function, ask: does this do exactly one thing? The next time you design a database table, ask: what new feature might we add in six months, and will this schema support it? These questions, asked consistently, are what distinguish the developers who move fast from the ones who plateau. What habit has made the biggest difference in the quality of your code? I'm curious to hear from engineers at different levels.

How do you know if your coding agent is actually helping? 💡 Moving beyond the hype, the real test of a coding agent lies in its measurable impact. It's not enough for it to "feel" faster; you need hard data to prove its value. In this video, we break down the 4 practical KPIs your team should track to measure the true contribution of your AI assistant: 📈 Time Saved per Ticket or PR: Is your team shipping code more efficiently? ⏱️ Code Review Velocity: Are pull requests getting merged faster? 🐞 Bug Catch Rate (Pre-Merge): Is code quality improving before it even hits production? 😊 Developer Satisfaction: Is the tool making life easier, or is it a source of frustration? Don't let your investment in AI be a shot in the dark. A data-driven approach allows you to see what's working, what's not, and how to optimize your workflow. Ready to see the numbers? Watch our latest episode now! Visit our website: https://nucleusls.com/ #AI #Coding #SoftwareDevelopment #TechMetrics #Productivity #Engineering #DataDriven #AIinDev #Metrics

6

1 Comment

Angela Montez's journey from analyzing complex texts to tackling Audible's extensive codebase exemplifies how diverse backgrounds can bring unexpected strengths to engineering. "My background in English lit actually helped since I'm used to analyzing and interpreting complex texts — now it's codebases instead of literature," she says. Explore how Angela's passion for connection and commitment to "Activate Caring" have fueled her growth from Audible intern to full-time Software Development Engineer: https://adbl.co/3WYwduk #InsideAudible

48

How to write the best prompts for a coding AGENT: Pair with another HUMAN on understanding and describing the problem. Still the same as before.

6

Build-in-Public Update | Day 30-1 Machinist's RAG AI Agent Yesterday's focus was testing how my PostgreSQL schema handles more complex engineering drawings. Loom Video The attached Loom captures the process of me processing a GD&T-heavy print through Werk24’s WebSocket/Python API using my Replit bridge app instead of my n8n workflow. The goal wasn’t to get better results, but to gather new sample data for schema validation. Costly Mistake Until now, most of my test data came from one manufacturer. That turned out to be a mistake. And a big oversight on my part. As a consequence, my schema was effectively tuned for that company’s drawing conventions, which meant my PostgreSQL DB didn’t generalize well. Fix To correct that, I’m collecting prints with a wider range of GD&T callouts and part types. Bigger Picture This is part of a broader process I’ve been working on over the past month: designing a database structure that can support multiple data types — part prints, 3D CAD files, CAM programs, setup sheets, and inspection reports — while remaining efficient for vector search.

2

After months of late nights, endless iterations, and more almost there moments than we can count, Levelop.dev (Beta) is finally live. Levelop began from a frustration that just wouldn’t fade. So many talented engineers lose momentum between interviews, not because they lack skill, but because staying consistent feels impossible. We wanted to change that. So we built a system that makes staying ready feel natural. 💻 Practice coding with instant feedback that teaches, not just grades. 🧠 Work on one real system design problem each week, from requirements to capacity planning to architecture. ✉️ And when you’re ready, reach real recruiters through a cold email pipeline that works in just a few clicks. Levelop isn’t another prep tool. It’s a rhythm. Learn, build, revise, and reach out, without burning out. 🌐 Try the beta → levelop.dev We’re just getting started and building this with the community that needs it most. If you’ve ever lost momentum between interviews, this is for you. #Levelop #BetaLaunch #BuildInPublic #StartupJourney #Coding #SystemDesign #CareerGrowth

6

📚🧱 Layout Engineers: Recruiters Notice This Resume Layer Stack Section But Few Explain It 🧱📚 Most resumes say “worked on layer stacks.” Recruiters glance. Then move on. Why? Because listing layers without context says nothing. Top engineers show how their stack decisions improved performance, manufacturability, or reliability. Here’s what strong resumes reveal first 👇 • They define stack purpose Power, ground, signal, shielding. Pro tip: Purpose beats listing. • They explain material choice Metal type, thickness, or dielectric. Pro tip: Choice signals expertise. • They connect stack to timing Critical nets and clocks considered. Pro tip: Timing linkage shows seniority. • They address EM/IR impact Current handling, voltage drop mitigation. Pro tip: Power-aware engineers stand out. • They show layer assignment strategy Which signals go where and why. Pro tip: Logic over arbitrary placement. • They mention cross-team alignment PD, verification, and fab teams synced. Pro tip: Collaboration matters. • They demonstrate iteration Adjustments across ECOs or design milestones. Pro tip: Iteration proves ownership. • They quantify outcomes Reduced noise, better signal integrity, cleaner signoff. Pro tip: Metrics validate effort. • They show constraint awareness Design rules, spacing, and thickness limits respected. Pro tip: Compliance elevates impact. • They link stack to final result Fewer violations, smoother tapeout, higher yield. Pro tip: Outcome closes the story. Weak line: “Managed layer stack.” Recruiter-ready line: “Defined and optimized metal and via layer stacks, aligned with PD and fab teams, improving signal integrity, mitigating IR drop, and ensuring clean tapeout.” Same work. Different signal. 🚫 STOP applying blindly! Follow us for real job-winning strategies. 🚀 Want daily job search hacks that actually get results? 👉 Follow JobLanderz.VIP for daily job hacks & insider tips! 📣 Follow JobLanderz.VIP now so you never miss a post! 👥 Tag a layout engineer refining their layer stack story! #LayoutEngineering #LayerStack #VLSI #PhysicalDesign #ASICDesignJobs #ResumeTips #JobLanderzVIP Extra tip: If your layer stack choices prevented late-stage rework or yield issues, highlight that explicitly.

It sounds romantic: "Software built by an inspector, for inspectors!" Here is the reality: Building software is expensive. Maintaining servers is hard. When the costs mount up, that "inspector-developer" stops updating the app and goes back to inspecting houses to pay the bills. You get left hanging with dead technology. Don't run your business on a hobby project. Run it on specialist-grade technology supported by world-class infrastructure (like Square). #HardTruths #InspectionTech #BusinessRisk #SwiftReporter #Specialist

1

Hiring Software Engineers? Write the role like a story, not a spec. Instead of: ❌ “We need someone with 5 years React experience” Try: ✅ “You’ll own the frontend across our learning platform, working directly with design and product to shape the student experience.” The best candidates don’t apply to job ads that read like chores. Want help refining yours? We do this daily. Re:Sourced

3

2 Comments

Welcome to Modifier Monday—your weekly dose of real-world guidance on the most misunderstood characters in coding. Modifiers can make or break a claim. Used correctly, they tell the full story of what happened during the encounter. Used incorrectly? They trigger denials, audits, and compliance risk. Throughout July and August we’re diving into some of the most common (and commonly misused) modifiers: ➡️ 25 – Significant, Separately Identifiable E/M ➡️ 59 – Distinct Procedural Service ➡️ 24 – Unrelated E/M During Post-Op ➡️58, 78 & 79 – Global Period Modifiers ➡️Payer Specific Modifiers and more…. We’ll break them down in plain language—what they mean, when to use them, and how to document with confidence. ✅ Practical examples ✅ Compliance tips ✅ Real talk about payer expectations Let’s clean up your modifier usage—one Monday at a time. #ModifierMonday #HealthcareInspired #MedicalCoding #ComplianceTips #AuditReady #RevenueCycle #DocumentationMatters #CodingEducation #HealthcareCompliance

15

3 Comments

Your coding agent will write tests. What it won't do, by default, is write good tests. Mutation testing fills that gap. And it turns out LLMs benefit from it far more than human developers ever did. Neal Lindsay breaks down exactly how to set it up and get your agent to run it on its own. https://lnkd.in/gvh_ndpQ

8

Good engineering starts with the truth. I once had a boss say, during a sales call: "if anyone asks, we're experts at socket programming in C#". We weren't; not by a longshot. And I wasn't comfortable lying about it, so I didn't. I still won't. The same goes for interviews. If I'm asked about a mistake I made, I won't downplay it; I'll lean in. I'll tell you about a *big* mistake I made - and the even bigger effort I put in to make it right. I don't want to land a job by pretending to be someone I'm not. And I certainly don't want to build systems on that kind of foundation. At this point in my life, success means feeling good when I wake up in the morning.

284

32 Comments