Coulton Vento

Protect Claude Code's Context. Thank Me Later.  I think of Claude Code's context window like a brain. And you can quietly destroy it by letting bad messages pile up. Garbage in, garbage out. Two tricks I use daily.                                                                                                                                                                                   Trick 1: ESC and Time Travel                                                                                 Short drifts are fine, let them play out. But the moment you notice Claude has been going the wrong direction for a while, hit ESC. Don't let it finish. Scroll back to your original message, rewrite it with more clarity, and this time add a hint about the exact wrong path it took. You just stopped a thread of bad thinking from living in your context permanently.      Trick 2: ESC, Distill, Clone, Resume                                                                             When debugging gets messy and Claude goes down five tangents before finding the answer, stop it. Ask it to distill everything into a markdown file. Clone that file so time travel won't erase it. Then resume before your original message and attach the file as context.                                                                                             Clean slate. Right answer. No tangent graveyard.                                        #ClaudeAI #VibeCoding #SoftwareEngineering #AITools #DeveloperProductivity #BuildInPublic

7

I built an iOS app with enterprise SSO and multiple integrations in under 6 hours this weekend. Not a toy. Not a tutorial project. A production-ready app based on existing business logic, with proper authentication. The tool: Anthropic's Opus 4.6, released last week. What's different? Tasks run longer. The model maintains context across complex, multi-step work without you babysitting every prompt. You describe intent, it executes. Six hours instead of six weeks. And here's what the "AI is killing software" crowd is missing: This doesn't replace developers. It compresses development. Faster builds don't mean fewer people. They mean more shipped. More experimentation. More ambitious roadmaps that were previously "too expensive to try." The $1 trillion selloff in software stocks this week was investors panicking that AI makes software obsolete. They've got it backwards. AI makes software easier to build. That means more of it, not less. The real gap isn't "can AI code?" — that question is settled. The gap is: can your team turn AI output into production? That's the skill that matters now. And most orgs don't have it yet.

47

19 Comments

Some Monday raw thoughts on why you might consider building smarter before you build faster, something it's getting more and more relevant in a world of coding agents

30

Google and Shopify just launched the Universal Commerce Protocol (UCP) — an open standard that lets AI agents discover and transact with any merchant. Target. Walmart. Etsy. Wayfair. Millions of Shopify merchants. All speaking the same language. I read through the spec and realized: there's no Go SDK yet. So I built one. And wrote a getting-started guide. https://lnkd.in/gfxKsfiq What the guide covers: → How discovery and capability negotiation work → Building a client that completes checkouts programmatically   → Implementing a UCP-compliant merchant server → Real, runnable code examples If you're building in commerce, payments, or AI agents — UCP is worth learning now. 📝 Tutorial: https://lnkd.in/grgq2FtQ #golang #opensource #ecommerce #ai #softwareengineering

37

1 Comment

AI will change frontend dev. Claude is great at UI, but still struggles with maintainable complex logic. Using SSR + jQuery/htmx lets Claude cook on the frontend while you focus on the backend. Designers vibecode the UI, backend devs own server logic.

16

4 Comments

Is it to late to have a career in web development? no, but you better adapt. Matthew Lawrence and I breakdown where we think the web development space is right now and give advice on how you can still break in. (Link on comments)

6

1 Comment

Building separate iOS and Android apps from day one sounds like the serious option. For many startups, it is just an expensive one. When the product is still proving its core loop, two codebases usually create cost before they create value: 1. More engineering time before launch 2. More QA surface on every release 3. More room for feature drift between platforms 4. Slower learning from real users The better early question is not 'native or Flutter?' It is: what does the business need to learn fastest? In many cases, a well-scoped cross-platform app wins because it helps the team: • launch earlier • keep one roadmap moving • reduce duplicate work • validate demand before overcommitting architecture Native absolutely makes sense for some products. But choosing it too early is often a status decision, not a business decision. If you are weighing native vs cross-platform for a new app, DM me. Happy to share how we would scope it.

1

“From vibe coding to agentic engineering.” Ronen Idrisov 🪄 is about to go on stage at TDX and this title says a lot about where things are heading. The last 12 months changed how software gets built: • less manual wiring • more intent, iteration, and feedback loops • agents moving from toy demos → real workflows What’s interesting isn’t just that AI helps you code faster. It’s that the abstraction layer itself is shifting. You don’t just write code anymore You define behavior, constraints, and systems that can act. Excited to see how the Salesforce ecosystem is thinking about this shift. If you’re at TDX, go watch this one.

58

Turns out getting money from VCs is actually very, very easy: Sign up for a Claude Code Pro/Max subscription. VC money is currently subsidizing hundreds (if not thousands) of $$ of your Claude Code usage per month. Cursor keeps shipping features to try to keep up with Claude, but there is no feature they can ship that will beat $200/mo for basically unlimited agentic coding. Here’s the thing though - the rug will get pulled on us eventually. That’s always been the plan, and everyone knows it. We just don’t know when the music will stop…or how many chairs there will be when it does. I’m worried it could get nasty. The gap will widen between the people/teams that can afford the tokens and those that can’t. But for now, the music continues - and my team is getting our money’s worth.

45

12 Comments

So...about cutting through the fancy math (AI) hype. I appreciate other folks making the problems clear: https://lnkd.in/eMw44ws9 It's one study, and it's small. Speed isn't the only measure. But it's pretty telling that people felt they were 20% faster with AI and were actually basically 20% slower. Great example of promise not matching reality AND confirmation bias hitting folks in assessment. And that's for use in software development, which is theoretically a pretty good use case! And some companies are catching on. The increasing tension between the optimism and the results is real. But a lot of companies still feel the pressure to "do AI" https://lnkd.in/e_hhFewA You can get real value out of machine learning for certain applications. GenAI can do some neat stuff. But a lot of people are learning the hard way that it just isn't magic. You've got to know what you're buying into. Whether you get it via my series (https://lnkd.in/eM9TYXRE) or somewhere else, a peek behind the AI curtain can help you avoid magical thinking

9

AI doesn't write the best code, even if your app works. Investing in AI code review tools is a necessity. They catch flaws that wouldn't have shown up until it actually mattered, like in production. For non-technical builders: when a review tool flags a bug or conflict you don't understand, paste it into your code editor's agent, Claude, or ChatGPT and ask it to explain the issue in plain English. These tools sometimes misread your intention as a bug, and if you don't understand what's being flagged and just let the agent fix it, you might be undoing something you actually needed. My favorite tools are Cursors review agent, BugBot, and Claude. #AItools #techadvice #codereview

19

2 Comments

The job description says "Software Engineer" but the role intake call tells a different story. I've had about a dozen conversations of this type in the last 18 months with early-stage SaaS founders and CTOs. They'll send through a pretty standard software engineering job description, but when we get on a call to discuss the role and what they actually need, something shifts. "I want someone who'll challenge the roadmap." "They need to be comfortable digging into usage data." "I need them to care about whether users actually benefit from what they build." "They should be asking questions about why we're building something, not just how." What they're describing isn't a traditional software engineer. It's a product engineer. Software engineers are often measured by output, clean code, shipping on time, stable systems - that sort of thing, Product engineers are measured by outcomes, user engagement, and whether the feature actually moved a metric. Both are valuable. But they're different roles that have different mindsets. If you're an early-stage company trying to find product-market fit, getting this distinction wrong means you'll either frustrate a brilliant software engineer who just wants to write great code, or miss out on a product-minded engineer who wants to own outcomes. What's been your experience hiring for these roles? #softwareengineer #software #product #productengineering

2

💭 Y Combinator CEO says YC startups launch with 95% of their code written by AI 😮. Harness Field CTO Martin Reynolds brings valuable insight to the conversation with LeadDev around this trend and its impact on software developers. "Having a human in the loop will continue to be vital. The future of coding will involve a combination of human creativity and AI-driven efficiency." Read more on how AI is reshaping engineering roles ⬇️. https://lnkd.in/gnRmrWxS

2

If you're frequently running out of tokens on Claude Code, switch to using Sonnet (type /model). Opus is 5x costlier for input, 8.3x for output, and you don't really need it except for occasional planning tasks.

8

99 Dev Problems with Rasmus Makwarth 99 Dev Problems features candid developer conversations on challenges, career growth, and tech trends. With insights, stories, and solutions from those in the field, it’s authentic tech talk for seasoned engineers and newcomers alike. Tune in for real-world coding and career advice. #99devproblems

3

Picking a dev shop shouldn’t leave the success of your startup to chance. But let’s be real: one wrong hire and you’re stuck: → Scope creep burning $50k/month while they "figure it out"… → Missed deadlines pushing your launch into next year… → "Scalable" code that faceplants at 1,000 users… You’re not crazy... 74% of startups blow their dev budget 2X–3X over (thanks, Gartner). Why? Most shops: ❌ Stuff your MVP with "nice-to-haves" (that users ignore). ❌ Lock you into tech that can’t grow with you. ❌ Smother you in jargon while your cash evaporates. We got tired of the BS. So we built a dev shop for founders, by founders: Tight-Fisted MVP Sprints Ship in 90 days with ONLY what makes you money. Live performance dashboards → see exactly how healthy your tech is. Skin-in-the-Game Pricing... Fixed monthly fees. Zero surprise invoices. We don’t "exit" ’til you’re winning. Why this works when others flop: While rivals drown in tech debt, you’ll: → Deploy 3X faster (clean code ≠ rocket science). → Pivot over coffee – not corporate Zoom marathons. → Scale profitably while others beg for "just 6 more weeks!" Stop rolling dice with your company. 👉 "DM" us for a free tech autopsy : We’ll dissect your stack + show you 3 cash leaks bleeding you dry. PS: Choose the perfect dev partner and watch your startup soar. #NoMoreTechRegrets #MVPThatDoesntSuck #RealTalkDevs #SaaSFounders #ScaleWithoutTears #DevShopRehab #StartupTruths

2

Evil Martians have reached a significant milestone with TutorialKit.rb: a toolkit for building interactive Ruby and Rails tutorials that run entirely in the browser, no installation required. The release candidate is ready for general use. The pitch is compelling: any Ruby gem maintainer can now create a polished, interactive tutorial using a single scaffolding command, deploy it to Netlify or Fly.io via generated CI workflows, and have learners writing real Rails code in the browser within seconds of arriving on the page. The engineering behind it is genuinely interesting. ruby.wasm runs inside WebContainers, HTTP requests are routed through a JavaScript bridge to reach external APIs, and a hybrid build strategy separates C extension gems (which require a full recompile) from pure Ruby gems (which can be repacked in seconds). The result is boot times of 2 to 5 seconds and a much faster development loop for tutorial authors. For the Rails community, the practical benefit is lowering the bar for learning. A new developer can work through a live Rails tutorial without touching their terminal. A gem maintainer can ship interactive documentation without asking users to set up a local environment first. Worth bookmarking if you maintain a gem or work in developer education. #RubyOnRails #Ruby #WebAssembly #RubyWasm #RailsDevelopers #DeveloperExperience #OpenSource

29

Red flags when evaluating offshore dev agencies 🚩 🚩 "We have developers in every tech stack" Translation: We'll say yes to anything and figure it out later 🚩 "Rates starting at $15/hour" Translation: You're getting mass-produced code from mass-produced coders 🚩 "We can start Monday" Translation: These devs have been on the bench for months. Ask why. 🚩 "100% satisfaction guaranteed" Translation: We have no legal definition of satisfaction 🚩 "We've worked with Fortune 500 companies" Translation: We did QA for a subsidiary's internal tool once 🚩 "Our developers work YOUR timezone" Translation: We'll burn them out in 3 months and swap them without telling you 🚩 "Full-stack developer who does it all" Translation: Master of nothing, debugger of everything 🚩 "We handle everything so you don't have to manage" Translation: You'll have zero visibility until something breaks 🚩 "Agile certified team" Translation: We do standups. Sometimes. 🚩 "We're not like other agencies" Translation: We are exactly like other agencies — The only question that matters: "Can I talk to 3 clients you've worked with for 12+ months?" Odesa.co What red flags did I miss? Drop them below 👇 #Hiring #OffshoreDevelopment #Startups #Outsourcing #RedFlags #techhiring

9

1 Comment