Mark Manley

I see so many startups claiming they want to hire elite Software Engineers, but then demanding a coding task before even having an intro call. That’s not how you attract top talent.

27

17 Comments

Coding agents are fast. But without real context, a lot of the code they generate is unmergeable. Wrong props, bogus states, subtle regressions—the usual AI slop. Storybook MCP fixes that. It turns your stories, docs, and tests into machine-readable context so agents can actually follow your patterns. Benchmarks show this produces higher-quality code 3× faster with 50% fewer tokens. And when agents get things wrong, MCP gives them a self-healing loop. They can run your interaction + accessibility tests, see what fails, and fix their own bugs before you ever open a PR. We’re assembling a group for Early Access to help shape MCP before launch. 👉 Want in? Request access here: https://hubs.li/Q03TCY9Y0

223

21 Comments

If you’re a US startup hiring mobile this quarter: screen for release ownership, not Flutter trivia. Senior candidates don’t separate “engineering” from “shipping.” They own releases, hotfixes, and the messy production stuff that actually matters. One question that works: “Tell me about a release that went sideways. What broke, how did you diagnose it, and what did you change so it wouldn’t happen again?” The answer tells you if they’ve actually owned production when it broke.

17

5 Comments

For years, we treated "good code" as optional. Tests? Nah, I have to ship. Types? I know what I'm doing. Docs? It's all in my head. It is counterintuitive, but agentic coding forces us to pay the "best practice tax" we've all been dodging for decades. It is now the most valuable debt your eng org can pay off. I write about some of the non-obvious things we do in our codebase to give agents superpowers, in today's Bits post (link in comments)

24

4 Comments

We're working with the US Navy to develop an intelligent tutoring system that improves the navigation, seamanship, and ship-handling skills of naval surface warfare officers (SWOs). KWYN® MARLINSPIKE enables #virtualtraining that allows SWOs to practice a wide range of scenarios, including navigating crowded waterways, to improve their ship-handling abilities. https://ow.ly/wBTq50Xrg9i #IITSEC #IITSEC25

15

Inspector (YC F25) is the AI IDE for front-end that lets developers prompt, edit, and build side-by-side with what their users see. Founders Quentin Romero Lauro and Michael Klikushin spent the past few years building tools and interfaces that sit between humans and computers — from HCI research at Carnegie Mellon to products at Character.AI and Oracle. When they started building web-based games, they ran into the same pain every front-end engineer knows: endless tweaking, context-switching, and debugging across Figma, Cursor, and Chrome DevTools. That’s how Inspector was born — an AI IDE for front-end that connects your browser and codebase. Inspector lets you select elements visually and prompt changes, giving coding agents access to the same runtime context developers have: visuals, elements, and logs. With Inspector, you’re editing right where things happen — in the front-end, not buried under tabs and screenshots. Congrats to the team on the launch! 🚀 https://lnkd.in/gV4bGFnd

598

80 Comments

Sharing a project I’ve been working on recently; a coding task tool to help companies hire software engineers faster. It's called: Task45.com As a recruiter of software engineers, I see the time it takes for companies to build and review coding tasks. A few days to send it out, a few more to find time to review and boom, the candidate has gone. With AI gaining trust, now might be the time to find a better way... Enter Task45 💥 With just a prompt (or a job description), Task45 will build out a coding task for you to share with engineers. As soon as the task is complete, it will review it instantly. Key benefits;  -45 min duration (to increase submission rates)  -AI-pair programming (for guidance and collaboration)   -Cheat-detection capability (for honest submissions)  -Immediate AI scoring (so you can act fast) -Human override options (for training the model) It’s ready to use today. Who's it for? -Recruiters -Non-technical founders -Time poor engineering teams If you think this could help shorten your hiring cycles, head to www.task45.com and give it a try! All feedback welcome! 

32

5 Comments

🚨 Censys on #React2Shell (CVE-2025-55182): We observe ~2.15M exposed web services running Next.js or other RSC-based frameworks—mostly in the U.S. and China. Not all are vulnerable, but active exploitation is underway. Patch now. 👉 Full advisory: https://hubs.ly/Q03X5QCS0

23

1 Comment

AI coding agents aren't just for coding anymore and if you're not paying attention you could be sleeping on one of the biggest shifts in personal productivity. I've been running Claude Code as my personal chief of staff. Not writing code. Managing my day. It schedules meetings. Prioritizes my tasks. Reviews Slack, Discord, and WhatsApp messages and surfaces what actually matters. It checks my calendar for conflicts, preps me for calls, and handles the operational overhead that used to eat my time. The shift isn't from to-do lists to better to-do lists. It's from to-do to to-done. The command line can seem intimidating but it's just a prompt, a place to type. What's different is that it can make tools on your behalf and then use them. - I wanted to transcribe a zoom call. Tool created. Done. - I wanted to update my website with notes. MCP installed.Done. - I wanted to get contact info from my CRM and airtable. Tool created. Done. Founders: the same agentic tools you're using to ship code can run your calendar, triage your inbox, and coordinate across the dozen communication channels you're drowning in. The early adopters aren't just coding faster. They're operating faster. What are you using AI agents for beyond code?

52

18 Comments

If you're a dev who's searching for a job, you may want to have your own professional email account. And the good news is - there's a way you can do it for free. In this guide, San shows you how to set up a custom email with Cloudflare and Mailgun. https://lnkd.in/geAm_F2u

637

27 Comments

I've been quiet on LinkedIn while we reimagined how Squidler works. Here's what we realized: Your coding agent knows what you wanted to build. It has the full code and context for every feature you prompted and every user story you described. But it not great att verifying, especially not at runtime. It's not design to run your app, or reason around it, as a user would. Cursor, Lovable, Claude Code, Replit and the rest, have made the Build phase magical. The amazing teams behind these products have pushed Building further and faster than I ever expected. But Building and Verifying are different jobs. This is why we redesign the navigation and interface of Squidler to close the loop between the two. Now, your coding agent can use our MCP integration to describe user goals, based on what it implemented, in the same language your agent already thinks in. Squidler then executes them, as a user in a browser, and answers three core questions: * Does it work? i.e. can users actually achieve their goals? * Is it broken? i.e. does it pass our rule based runtime validation? * Is it easy? i.e. was is intuitive or hard to achieve the goals? Goal-driven. Runtime-verified. Your agent knows what you wanted, and with Squidler you can now check that you got it. Check the comments for demo and a link to try it.

63

18 Comments

How to 2x your eng team output w/ coding agents: make an overnight "Accelerated Task Pipeline." 1. Set up a coding agent in the cloud w/ access to your repos ✅ (We use our own Coworker.ai agent called Cloud Merlin, but use whatever you prefer.) 2. Every JIRA ticket that's categorized as "Accelerated" -- i.e. bugs or task-sized features -- automatically gets fed to cloud agent ✅ (KEY NOTE: the relative cost of trying to "fix" or "build" something w/ AI is so low, we don't check the validity of the report until AFTER the AI comes up with an attempted solution overnight. Think of this as a sort of async processing for organizations, optimizing for speed.) 3. Rotating daily engineer reviews (3a) quality of bug report/feature requirements (before reviewing code). If not good enough, kill draft PR and send back to reporter (3b) if requirements (basically the "prompt" in this case) are good enough, review the code itself and go through all security/privacy checks etc. ✅ 4. Ship it in that day's release ✅ The onus of "thinking through" a bug or minor feature shifts from an individual engineer -> ops+product+AI. Massively cuts total time spent investigating and speeds up feedback loop from customers/ops -> product. Goodluck

31

5 Comments

SOTA models hit 67% on coding benchmarks. That sounds good until you realize the 33% don’t fail quietly, they can fail expensively, leading engineers down rabbit holes and burning hours. This is the productivity lottery that the benchmarks miss. We ran Gemini 2.5 Pro, Claude Sonnet 4, and GPT-5 on SWE-bench with professional engineers analyzing every failure. What we found: Small mistakes spiral into self-reinforcing hallucinations. Some models can’t backtrack. Others succeed by noticing uncertainty and re-checking instead of guessing. Leaderboards don’t show you this. Trajectories do. Full breakdown here 👉 https://lnkd.in/euEShCZZ

24

1 Comment

When coding agents fail to accomplish what we task them with, very often it's because the codebase has complications we failed to anticipate and communicate. It's easy to get mad at the AI, but reckoning with technical debt requires your awareness and involvement, too.

11

We just released Taro's newest course about what life is really like at FAANG. So much of the discussion about working at FAANG focuses on things like the scale of the work or the free snacks. But on the day-to-day level, an engineer at Google or Facebook doesn't think about these topics. They are spending most of their waking hours on things like: - how much scope do they have? - how much control do they have vs just being a cog in the machine - what do their coworkers think about them? As a 5 year veteran of Meta, I break down the real pros and cons of working in Big Tech. Check it out here: https://lnkd.in/gfkwVDf9

58

4 Comments

The fastest growing state for software engineer salary submissions is Florida with a median income of $125k. (Levels.fyi, EOY report). I wouldn’t have guessed Florida. I'm always a little surprised when I come across a profile or resume that has FL in the address. But having spent a lot of time there, here’s why it makes sense: - There's no state income tax, which for high earners can be a big deal - You can drive from rural farming towns to the beach in just a couple hours - There's no snow - The gulf seafood is hard to beat What’s interesting isn’t that it's Florida, it’s the bigger picture trend. More and more people don't want to live in tech hubs, they're optimizing for lifestyle + income + flexibility. Remote work didn’t just change where people work. It changed what they’re willing to tolerate to keep working. If you could keep your current salary and pick any state to live in, where would you choose?

51

12 Comments