Becca Barton

The Real Cost of Hiring the Wrong Engineering Team Good engineers build what you ask for. Great engineers question what you missed. There’s a myth in product circles that shipping is just about speed. That if you assemble a fast-moving team with the right stack, success follows. I’ve seen what happens when that’s the only filter. One client came to us after sinking six months and six figures into a dev shop that “checked all the boxes.” The result? A barely working MVP, hardcoded APIs, and infrastructure that collapsed under basic traffic spikes. Their mistake wasn’t tech selection. It was team alignment. At Perisync, we’ve spent six years building a virtual engineering model that prioritizes: ➤ Context fluency: Our devs don’t just know React or Node—they understand the product layer. ➤ Asynchronous rigor: Remote teams can’t rely on heroics. They rely on clarity. ➤ Accountability loops: We embed into client PM cycles. No “throw it over the wall” delivery. And here’s the part most teams miss: It’s not just about skill. It’s about thinking like an owner. When you staff based on headcount, you get hours. When you staff based on ownership, you get outcomes. If your team is burning time fixing what they built last month—it might be time to rethink who you’re building with. Let me know if you want to see how our "hire-train-manage" system works across borders without losing velocity. #RemoteEngineering #TeamScaling #TechHiring #VirtualTeams #SoftwareDelivery

314

5 Comments

I'm building a go-to-market team of 0. At Remotely, AI already powers our workflows across the org. This is about letting it own the job. The product is simple. What i'm actually testing is whether an AI agent can do the job of an employee. Can it run go-to-market by finding, qualifying, and delivering value to customers without a human in the loop? Not inside Remotely.works since we have investors, a team, clients who depend on us. The cost of getting it wrong is high. So my co-founder and i built a side project instead (he's also named Pau, which confuses people). Low stakes, high curiosity. It's called Dealhound. I bought a Mac mini and set up OpenClaw to try. Not a human using AI as a tool but AI running a business function on its own. One thing i didn't expect? Because the agent checks prices daily, it's accumulating real pricing data over time. Daily checks so you actually know when a 30% discount is a deal versus just marketing. That wasn't the plan, but it might be the most interesting part. I don't know if Dealhound becomes a real business. That's not really the point. The question underneath it is... what can an AI employee actually do today, and where does it break? Easier to find out with a side project where the downside is a Mac mini collecting dust than inside a company where people are counting on you.

81

9 Comments

Having coffee with YC founders and indie hackers in San Francisco taught me one thing. Nobody is obsessed with features. They’re obsessed with removing pain. Every conversation comes back to the same themes: On-call burnout. Repetitive operational work. Being paged at 2 AM for problems that follow patterns. What stood out most was the mindset. No ego. No flexing. Just curiosity. The best builders aren’t asking, “How do I build more?” They’re asking, “What can we delete?” Delete boring work. Delete constant alerts. Delete the need for humans to babysit systems. That’s where real leverage lives. SF doesn’t reward noise. It rewards founders who listen deeply and build what shouldn’t exist anymore.

24

11 Comments

Most TA teams are hiring for roles that no longer exist. Jenny Wen, Head of Design at Anthropic, said something this week that every TA leader in tech should read. Design craft that used to be 60-70% of the job is now 30-40%. The rest is shipping in code, pairing with engineers, steering direction. In software engineering, AI is writing 50-90% of code at leading tech companies. The execution layer is gone. What's left is judgment and adaptability. But your job descriptions haven't changed. Your scorecards haven't changed. You're still hiring for the execution layer. Three things to fix now: 1. Requalify your senior briefs. Experience in an obsoleting process is a risk, not an asset. 2. Take blank slates seriously. Fast learners with no process baggage are outperforming seasoned practitioners at the sharpest companies right now. 3. Rebuild your assessment criteria. If you're testing for the old workflow, you're screening out the people you actually need. The best hire 18 months from now doesn't look like the best hire today. Is your process seeing that?

112

5 Comments

Woah, LinkedIn is filled with YC and vibe coding. Why are they pushing it so much? Turns out, it's not just hype, it's capital 😉 they haven't launched a "vibe fund", but they’ve quietly backed dozens of AI-native startups through their standard $500k checks. That’s tens of millions indirectly fueling vibe-coded teams. Nearly 25% of startups in recent batches are building with 95%+ of their code written by AI. Teams of 10 doing the work of 50. Some pulling $1M–$10M ARR. Solar (Lumenary) is one example. But it's bigger than one startup, it’s a pattern. Vibe coding isn’t a trend. It’s becoming a YC lens on the future of product velocity. Still early. But the signal is strong.

19

4 Comments

Lots of people talking about harnesses, but what your org really needs it its own coding agent like Ramp, Stripe, and Coinbase have built An internal Claude Code built for every technical and on technical employee at your company My article below cuts through all the noise to clearly lay out all the decisions your org needs to build its own coding agent.

14

1 Comment

I've been thinking a lot lately about how we've adapted to AI at PocketSeed, and how being a small team has let us really leverage it and shift our workflows quickly. Traditional roles are blurring, designers have direct access to codebases, and engineers are moving faster and getting more creative with what they explore. It's completely changed the way we design, build, test, validate and ship products. Alongside that, it's opened up whole new worlds for creators and builders, giving us access to explore more complex ideas and projects without needing a big team or outside help. Lately in my spare time I've picked up electronics and robotics. I'm 3D printing arms and cases, wiring up microcontrollers, and integrating services and AI into them. My latest project is a desktop assistant with access to multiple local and cloud models for different tasks. It has various personalities and skills. It's based on OpenClaw, but I decided to rebuild the platform from scratch to understand how the pieces connect and the strengths and downfalls of different approaches. This weekend I gave DOT-001 a mechanical arm to play with. It can be fully controlled by the assistant, or by hand gestures through a webcam in the browser. Building something like this used to feel completely out of my realm, like pure sci-fi, the kind of thing only kids studying engineering at expensive universities got to do. But I spent 50 bucks on Amazon, 3D printed some parts, and a couple of weekends messing around and chatting with AI to get it working. A small breakdown: 2x ESP32 boards, one for the arm and one for DOT-001 OLED and TFT touch screens for displays PIR sensor for detecting people nearby Claude API and Ollama (Qwen3 4B) for cloud and local LLMs Custom personalities and skills, including a replica of myself with a cloned voice (which is a little weird if I am being honest. 😅 ) Voice via ElevenLabs API Arduino, Python and React for code and interfaces MediaPipe for movement tracking Grippy Bot by Roboteurs from Thingiverse for the arm design Custom designed enclosure for DOT-001 The sheer possibilities opened up to creators and builders is incredible. It gives me so much hope for all the cool and useful things people will build now that they have access to these tools. If you're curious about any of it or have questions, leave a comment and tell me what you're building!

18

3 Comments

A big moment for us at Croissant! We raised $28M to keep building a platform that helps people shop smarter with ... Real product discovery through a community feed of authentic shopping activity. 10% back when you shop partner brands using Croissant. New financing tools that help brands grow. Really proud of this team. A lot of hard work went into getting here. No single breakthrough. Just showing up, listening, and stacking bricks. This isn’t the finish line though, it’s the starting gun! Time to get back to work. If you haven’t yet, check out the new Croissant app: https://lnkd.in/g8p4FYHm

39

5 Comments

🚀 Hot take: React is great.   But *Next.js?* That’s what serious builders pick when *outcomes* matter. Let’s be real — we’re not just building UI anymore.   We’re building entire ecosystems.   And in that game, **Next.js isn't just a framework — it's an unfair advantage.** Here’s why Next.js dominates every frontend decision I make (B2B, B2C, SaaS, dashboards — you name it): 💡 **1. The Missing Intelligence Layer**   React is a library.   Next.js is a full-stack system with routing, SSR, file-based APIs, static generation, and server functions.   Your app isn’t stitched together from 10 npm packages — it just *works*. You ship faster, debug less, and avoid Frankenstein architectures. 🧠 **2. Built-in Performance Engineering**   Next.js comes with smart defaults like: - Image optimization   - Code splitting   - Lazy loading   - Hybrid rendering (SSR + SSG)  So you don’t just build faster — your *users* feel the speed. 🔍 **3. Error Tracing and Observability = God Mode**   Next.js + Vercel gives unmatched visibility into: - Rendering bottlenecks   - API response time   - Route-level analytics   - Deployment rollbacks  It’s not just logs.   You see the problem **before your user does**. ⚙️ **4. Backend Capabilities, Frontend Simplicity**   Need a database call in your page?   Drop it in a `getServerSideProps`.   Want a real-time API?   Spin up an edge function or serverless route. No need for a separate Node.js backend.   Your entire stack lives in one repo, one mental model. 🔐 **5. Production-Ready Out of the Box**   With React, you need to configure webpack, Babel, routing, error boundaries, and more.   With Next.js, you’re already there.   It’s battle-tested at scale — from Shopify to Notion. 📦 **6. Edge Deployment + ISR (Incremental Static Regeneration)**   This is *pure gold* for B2C and SEO-heavy apps.   Next.js doesn’t rebuild your entire site.   It updates only the pages that change.   Blazing fast delivery + dynamic freshness. Win-win. 🔥 TL;DR: If you're building serious web apps,   👉 Next.js isn’t *just* better than React.   It’s a **thinking framework** for engineers who play to win. It removes the “plumbing” decisions.   And gives you **clarity, control, and compounding velocity**. So the next time someone asks:   “React or Next.js?” Tell them this:   **React is a library.   Next.js is a weapon.** 👇 Are you team React or team Next.js? Drop your stack preference in the comments & tell me what you’re building! Follow me — Prasanth Kumar — for more high-performance product and dev insights ⚡️ #Nextjs #WebDev #FrontendEngineering #SaaS #B2B #ReactJS #ProductVelocity #DevTools #TechLeadership #Vercel #JavaScript #ModernWeb

5

Most teams underestimate what it takes to build real collaboration. I learned this the hard way at InVision. 18 months. 8 engineers. A collaboration tool that… barely worked. Meanwhile, Figma shipped seamless multiplayer and took the market. That experience burned one truth into my brain: → Every product will need collaboration. → Most teams will face the same nightmare we did. So Guillaume and I walked away from safe jobs to solve it. At first, we chased the wrong idea: a creative tool. But every month, the real-time infrastructure pulled us back. It was clear the “side problem” was the actual problem. That’s when we flipped everything and built Liveblocks. Phase 1: infrastructure for developers. Phase 2: pre-built React components (Google Docs-like collaboration in minutes, not months). Now Phase 3: Liveblocks 3.0 embedding AI copilots so humans and AI work together like teammates. Because the real unlock with AI is collaboration. I talked about this on Founders by Earleads podcast: → Why 95% of enterprise AI implementations fail → The difference between AI-native vs. bolted-on experiences → Why existing companies have an edge over AI-first startups Links in the comments. Thanks Othmane Khadri for having me. Every detour led us here. The future belongs to companies that make AI feel like a teammate, not a tool.

31

4 Comments

The whole Liftoff team was in New York last week for a team onsite. What a privilege. - We talked openly about how our product process has shifted. We’re now building 1 week - 2 month roadmaps instead of 3-6 months, similar to what we’ve heard from most teams we talk to. - Every discipline on our team beyond engineering – GTM, Design – is now shipping code. We’ve always been high-intention with what we ship and rapidly iterative based on user feedback. Now we’re scaling that intentionality and responsibility for product coherence throughout our team. That doesn’t just look like rapid shipping, it means tackling really hard, holistic product and systems problems together, so that everyone holds the big picture. Everyone needs to own product strategy, so that everyone builds things that work. - We’re a hybrid team that’s been deliberate about every hire, and that’s worked well for us. There’s a lot of talk these days about the amount of time people work, but not enough about the potency of that time. With a small team where everyone is switched-on, and shaping strategy – every minute feels productive. What I'm most grateful for is how much this team enjoys the actual work, and each other. We’re building as fast as we can, and hiring across all disciplines right now: customer success, GTM, engineering and design. Link in comments!

128

5 Comments

Cursor isn't just for engineers anymore. Stripe recently rolled it out to 3,000+ engineers but with vibe coding now part of the interview process at many companies, Cursor is quietly becoming the go-to tool for non-engineers, too. Amplitude's PM Frank Lee says Cursor is now firmly part of his analytics workflow thanks to their new plugin released last week: "with the amplitude plugin, i can pull product context directly into cursor to analyze dashboards and synthesize customer feedback, then push requests to cursor agents to draft PRs... call it the first self-improving product loop" We’re watching the traditional IDE turn into a product operating system. ◾Here’s some of the ways you can use Cursor for non-engineering workflows: 1. Build a self-running analytics pipeline. Install three of Cursor’s new Plugins -  Amplitude, Linear, Notion - and write one master prompt. From that point, a single command monitors your key metrics, flags anomalies against healthy ranges you define, auto-creates investigation tickets in Linear, and updates your Notion doc with findings. No need to distract engineers. 2. Understand parts of your codebase in minutes. Ask Cursor to generate an architectural diagram of your entire app - or just one specific feature. Then ask it to explain what each part means in plain English. Genuinely useful when you've just joined a new team and need a quick overview of the basics or shipped something you don't fully understand. 3. Test features without leaving your editor. Cursor's built-in Browser gives the AI direct control over a live version of your product - it can click, scroll, take screenshots and monitor network traffic. Useful for accessibility checks and spotting UX issues before they reach users. 4. Push live UI directly from Cursor and Claude Code into Figma. This is the one most people haven't used just yet. Released a few days ago, Claude Code + Figma's new MCP integration is now bidirectional - meaning you can capture a running web UI and push it straight onto your Figma canvas as fully editable frames with real layers and components. Not a flattened screenshot. Actual design files you can work with. 🔗Full step-by-step walkthrough of every use case, including how to set up the analytics pipeline, prompts you can use and the Figma integration from scratch here:  https://lnkd.in/ewnUR9Uu 

49

2 Comments

This market has been brutal to a lot of folks in tech given layoffs. After mentoring tons of junior level engineers, designers, product managers, here are some guiding questions I have given to others who were successfully hired that will help you rethink your strategy and approach. As we go into summer break to recalibrate, it's important reflect deeply on what does and doesn't work: 1 / Tracking Goals - Are you tracking your goals every single week? Every week, like YC does for Startup School (indicate what % you have grown your number of users etc.), add to your spreadsheet/table to tally your numbers - how many jobs you applied for, how many referrals you have (warm as opposed to cold). The most comprehensive dashboard I had seen was from another friend who tracked his leetcode (custom dashboard) that showed his practice so he could analyze what questions he was failing at where, why, and when, to optimize the speed and code quality for his interviews. Optimize. Remember you can't grow where you don't measure and can't see. 2/ Accountability groups - Do you have a consistent and regular support system? The most effective groups I joined met every week, followed a metric similar to above, and had an accountability group to report at least 1 target goal every week they wished to hit. When you stop going every week, you might see your progress wane. 3/ Your brand - More than hitting numbers Have you thought about how you're reflecting out what you want in the world? I had to take a hard look at what I was putting out there. Too much of my portfolio was dated (finally about to release a new one and a new app), and the marketing and SEO for this had led to many jobs I've had to re-examine the language even in on my LinkedIn profile. The moment I did that I attracted better opportunities and stopped getting the ones that didn't align. My resume either read TPM or LinkedIn read dev in areas I didn't focus on as my core. 4/ Internal Evaluation - Have you thought about internally on what fulfills you? Reflecting on my last company, I despised dealing with non-technical problems that were more of a headache and my own technical skills atrophied because I delegated so much that I wasn't hands-on enough on the product. Any time I have deferred anything to a technical or non-technical PM that doesn't have their hands somewhere in the code base ends up a recipe for disaster. It also just makes me really unfulfilled, empty, and unhappy like I haven't contributed anything substantial. This leads you to different roles you may consider and why (for me maybe more IC/tech lead than managerial). 5 / Are you doing all that you can do to find that job? If you go through #1-4 can you honestly say that you have tried everything? When the answer is no, then you may find indicators on why you may not be reaching your goal or finding alignment in your job search to be successful. #techjobs #SWE #softwareengineers #designers #PMs #designers #tech #jobs

17

1 Comment

I don't think software engineering jobs are going anywhere - I've seen proof multiple times in the last few weeks. I routinely jump on calls with people who complain "I'm using Claude to do web scraping for 'X', but it's really expensive. Can your product help?" I then ask what they're doing, and they're always really proud to show me (which I appreciate!): 1. They give Claude access to their laptop 2. It then uses their browser to scrape sites that are blocked (e.g. reddit) 3. Each page is like a million tokens of raw html 4. It does each page 1 by 1 I then show them what happens when you use a dedicated tool to scrape a webpage - the cost is IMMEDIATELY down 98%. Then you add in the fact that an open source model is just as good at scraping - that brings the cost down another 90%. Lastly... it's much faster if you can parallelize it. And no, not doing batches of 5 at a time. Doing 1000 at a time. We took a customer who was scraping reddit and LinkedIn for ~$0.90 per page, and got the cost down to half a cent per page.

5

A Director of Product friend at a Series C sent me their offer letter last week. They were ready to sign. I read it twice and sent them back 5 questions: The numbers were strong. The structure underneath was missing the protections that decide whether a senior offer actually pays out. The 5 questions I sent: 1. What is the exercise window on the options. Default is 90 days after departure, but 25% of startups and growing do 1-10 years. Anything less than 3-7 years for a Director-level grant is leaving real money exposed if you ever leave. 67% of in-the-money options grants expire unexercised (Carta report). Don’t be in that group. 2. Is the strike price guaranteed? They are in the middle of raising their Series D but quoting pre-round strike price. Your exercise costs may 4x if you don’t pre-negotiate a pre-round strike or at least guarantee a net $ spread between the new preferred and strike price to value your stock options. 3. What is the severance package. Specifically: months of base, continued vesting through severance, and COBRA. Standard for the role is 4 to 6 months base, continued vesting, full COBRA. A blank where severance should be means zero severance if let go on day 91. 4. What is the equity acceleration on change of control? The company is rumored to be in early acquisition talks. With no acceleration, if the deal closes inside a year, most of the equity goes back to the cap table and you gave up a safe bet to help others win while you get nothing. 5. What is the clawback on the sign-on bonus, and is it waivable for involuntary termination? The standard 12-month clawback is fine. The standard "no waiver for any termination reason" is not. My friend asked 4 of the 5 at the offer letter close. 3 were granted. 1 was negotiated to a partial concession. The headline cash did not move. The protected value of the offer moved by a meaningful amount. We help negotiation clients with these terms all the time. You'd be surprised how easy they are to negotiate if you ask. But how costly they are if you don't. This is the second negotiation that candidates don't know exists. How many of these 5 does your current offer answer cleanly?

40

6 Comments

Most prompt engineering advice is actively making your agents worse. The worst offender: telling your agent who to be. "You are a senior frontend engineer with 10 years of experience..." was only useful when the models were weaker. Role-playing helped them stay in their lane. Frontier models don't have lanes. They've already absorbed every senior frontend engineer's body of work on the open internet. When you tell one to "be" a senior engineer, you're taking a model with access to all of it and constraining it to the cliché version of one. Our agents got better when we stopped doing this. The new prompt looks more like a director's note than a job description. We describe the task and the constraints, and we let the agent figure out the approach. We don't tell it who to be. The output stopped sounding like the cliché version of a senior engineer and started solving the actual problem. If your agent prompts still open with "You are a..." you're 18 months behind.

10

2 Comments

"Focusing on the product is rarely wrong" - h/t Sarah Guo. "It took Dylan Field and Evan Wallace more than four years to go from founding Figma to public launch. They built Figma playing for the endgame, not the headlines. Dylan didn't chase growth at all costs, or settle for what they could ship in 10 weeks. He didn't try to crush competitors. Instead, he spent a decade obsessing over building the product he imagined should exist and the best company he could—making design and software product delivery more powerful and collaborative. He consistently acted with integrity and thoughtfulness. Turns out that's not just good philosophy. It's a strategy for a special business. In 2014, when John Lilly led Greylock's Series A investment, Dylan and his team had seven people in a nondescript office on Alma Street in Palo Alto, and barely an alpha product. But they described a future with remarkable foresight: a high-performance, collaborative, multiplayer design tool built natively for the web, harnessing all its power, changing product development forever. Figma had its fair share of challenges. Market size was an early question. $800K monthly burn pre-launch was terrifying. Like with any startup, getting early pilot customers required cajoling of our friends. Competitors who built less product, for a while seemed ahead. Yet Dylan persisted, driven more by conviction than fear."

24

2 Comments

Spoke with Head of Product at a large company (30+ PMs). They actively ship code and their expectation is that all PMs on their team do this. Their recommendation for the first 3 things that PMs and designers (and marketing!) should own end to end: 1. Public-facing website 2. Support experience 3. Onboarding / growth flows These three aren't random. - Website: you feel what a potential customer sees before they trust you. - Support: you feel what a customer experiences when you've failed them. - Onboarding: you feel where customer curiosity turns into confusion or commitment. All three are the seams where internal product meets customer reality.

534

36 Comments

If you're #hiring engineers, who’s selling your opportunity? I don’t mean pitching comp or perks. I mean treating candidates like humans from the very first touch point. Making them feel seen, respected, and genuinely excited about joining your team. Too often, I see founders, hiring managers, and even recruiters assume their opportunity speaks for itself... That great candidates will simply apply, get screened, and accept an offer. It’s a transactional mindset as if you're ordering a product, not inviting a whole damn person to build something with you. One of the most common patterns I see: The CTO shows up at the final round thinking they can “close” someone in 30 minutes—after the rest of the process has been purely evaluative. I once had to remind a client *three* times that their engineers weren’t even introducing themselves at the start of technical screens. This isn’t about incompetence or bad intent. Most hiring managers are just busy with more than a full time job, and unaware of how their process feels from the candidate side. Often times I'm asking very critical questions that effect *tone* that creates emotional reactions. . . - If a candidate has two competing offers, are you willing to wait a week to extend yours or can you close them today? - If you add an interview step mid-process, what message does that send about how you operate? - If your company shifts structure mid-process, are you willing to bet all-in on honesty and transparency or risk them smelling the dissemblance... or worse, quitting in 3 mos when they feel lied to? These moments don’t just win candidates—they build long-term trust and reputation. It’s not hard to avoid being a horror story in someone’s job search. But it does take intention, consistency, and someone who knows how to design a process that sells, not just screens. And no... There isn't an AI tool for this. That’s what I’ve done for 15+ years with early-stage and high-growth #startups like Tinder, Honey, Fender Digital, and Signal Sciences. If are interested in a noise-free pipeline of engaged, qualified candidates who are actually excited to solve your problems—we should talk.

7