About
I build software that reduces operational work, makes complex workflows auditable, and…
Services
Activity
2K followers
-
Scott Spence reposted thisScott Spence reposted thisWhat happens in Vegas doesn’t always stay in Vegas… Our colleague Nemo recently attended DEF CON, one of the world’s largest cybersecurity conferences. He put together this short video sharing his experience, what caught his attention and what he brought back to the Cloud Lobsters team. Take a look 👇 Thanks to Claire Whitaker at Sequentially for the video production. Scott Meechan Tara Whitaker Duncan Elliot Nemo Godebski-Pedersen Scott Spence Yehen Singankutta Arachchilage #DEFCON #Cybersecurity #CloudLobsters #Insurance #insurtech #artificialintelligence
-
Scott Spence reposted thisOne thing I have learned over the years is that the best engineers rarely need convincing. They want interesting problems, ownership and talented people to work alongside. That's exactly the culture we are building at Cloud Lobsters. If Tara's post resonates with you, we would love to hear from you.Scott Spence reposted thisOur strongest engineering hires usually found us first. So instead of asking where to find our next engineer, we asked ourselves a different question: What makes Cloud Lobsters the sort of place engineers enjoy working? We're an InsurTech company at heart, and that's where we've built deep expertise. Insurance is full of complex data, workflow and automation problems. But good engineering isn't tied to a single industry. Last year alone our engineers worked on AI-powered underwriting, banking workflows, logistics software and industrial optimisation. The domain changes. The engineering challenge is what keeps it interesting. That's reflected in how we build software: • 𝗣𝗿𝗼𝗯𝗹𝗲𝗺 𝗳𝗶𝗿𝘀𝘁, 𝗰𝗼𝗱𝗲 𝘀𝗲𝗰𝗼𝗻𝗱: We work directly with clients from day one. We spend time understanding the business problem, challenging assumptions, prototyping quickly and then building the right solution. • 𝗣𝗿𝗮𝗴𝗺𝗮𝘁𝗶𝗰 𝘁𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆: We use Svelte extensively, embrace AI where it genuinely adds value and choose technology because it solves the problem—not because it's fashionable. • 𝗕𝘂𝗶𝗹𝗱 𝘄𝗲𝗹𝗹, 𝗿𝗲𝘂𝘀𝗲 𝗼𝗳𝘁𝗲𝗻: When we create something genuinely reusable, it becomes part of our toolkit. Every project leaves us with better tools, so we spend more time solving new problems instead of rewriting old ones. • 𝗙𝗮𝘀𝘁-𝗺𝗼𝘃𝗶𝗻𝗴 𝗽𝗿𝗼𝗷𝗲𝗰𝘁𝘀: Most engagements last weeks or months rather than years, so you'll work across different industries, solve a wide variety of problems and see the impact of what you build. We keep our teams small, experienced and collaborative. You'll work alongside engineers who challenge each other, share what they know and care about building great software. Everyone contributes, everyone learns and everyone ships. We're growing our engineering team in London. If you enjoy owning problems from the first client conversation through to production, you'll probably feel at home here. You'll probably enjoy working with us if you: • enjoy understanding the business before writing the first line of code • care about writing clean, maintainable software • enjoy ownership, direct client interaction and project variety • want to work together in our London office If that sounds like you, drop me a message. I'd love to have a conversation.
-
Scott Spence reposted thisScott Spence reposted thisOur strongest engineering hires usually found us first. So instead of asking where to find our next engineer, we asked ourselves a different question: What makes Cloud Lobsters the sort of place engineers enjoy working? We're an InsurTech company at heart, and that's where we've built deep expertise. Insurance is full of complex data, workflow and automation problems. But good engineering isn't tied to a single industry. Last year alone our engineers worked on AI-powered underwriting, banking workflows, logistics software and industrial optimisation. The domain changes. The engineering challenge is what keeps it interesting. That's reflected in how we build software: • 𝗣𝗿𝗼𝗯𝗹𝗲𝗺 𝗳𝗶𝗿𝘀𝘁, 𝗰𝗼𝗱𝗲 𝘀𝗲𝗰𝗼𝗻𝗱: We work directly with clients from day one. We spend time understanding the business problem, challenging assumptions, prototyping quickly and then building the right solution. • 𝗣𝗿𝗮𝗴𝗺𝗮𝘁𝗶𝗰 𝘁𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆: We use Svelte extensively, embrace AI where it genuinely adds value and choose technology because it solves the problem—not because it's fashionable. • 𝗕𝘂𝗶𝗹𝗱 𝘄𝗲𝗹𝗹, 𝗿𝗲𝘂𝘀𝗲 𝗼𝗳𝘁𝗲𝗻: When we create something genuinely reusable, it becomes part of our toolkit. Every project leaves us with better tools, so we spend more time solving new problems instead of rewriting old ones. • 𝗙𝗮𝘀𝘁-𝗺𝗼𝘃𝗶𝗻𝗴 𝗽𝗿𝗼𝗷𝗲𝗰𝘁𝘀: Most engagements last weeks or months rather than years, so you'll work across different industries, solve a wide variety of problems and see the impact of what you build. We keep our teams small, experienced and collaborative. You'll work alongside engineers who challenge each other, share what they know and care about building great software. Everyone contributes, everyone learns and everyone ships. We're growing our engineering team in London. If you enjoy owning problems from the first client conversation through to production, you'll probably feel at home here. You'll probably enjoy working with us if you: • enjoy understanding the business before writing the first line of code • care about writing clean, maintainable software • enjoy ownership, direct client interaction and project variety • want to work together in our London office If that sounds like you, drop me a message. I'd love to have a conversation.
-
Scott Spence shared thisFive agents fighting over the same files is not scale. It is a coordination problem you built on purpose. Multi-agent setups are everywhere on my feed right now, so here is the line I keep coming back to in my own workflow: Use more agents only when the work is actually parallel. One agent researching an API while another maps the existing implementation? Genuinely parallel. Separate worktrees with a lead reviewing the results? Works. Five agents needing the same files, context and decisions? You have created overhead and called it velocity. And the part nobody wants to hear: the lead still has to understand the combined result. Delegation does not transfer accountability to a swarm-shaped cloud. I wrote up my full workflow, from harnesses to evals to secret handling. Link in the comments.
-
Scott Spence reposted thisScott Spence reposted thisWhat’s it like working at Cloud Lobsters? 🦞 For me, it’s using AI to rethink complex problems, building systems our clients love, and doing it alongside people I genuinely enjoy working with. I can’t imagine a better place to work - and we’re looking for more people to join us. Cloud Lobsters is a London-based Insurtech startup, and we’re hiring software engineers who want the chance to shape what we build. Curious? Take a look at life at Cloud Lobsters - and if it feels like your kind of place, drop me a line 👇 Thanks Claire Whitaker at Sequentially for the video production! 🙏 Tara Whitaker Duncan Elliot Scott Meechan Scott Spence Nemo Godebski-Pedersen Yehen Singankutta Arachchilage
-
Scott Spence posted thisMe as a self employed individual and my ‘titles’… Fractional CEO Fractional CTO Fractional CMO Fractional Developer Relations Fractional Tax Analyst 🥲
-
Scott Spence reposted thisMy sabbatical is over (so much for me taking a break from the web!) As the summer starts to fade and stakeholders are back in office, do let them know if they're on the hunt for a lead/principal frontend engineer, do give me a shout.
-
Scott Spence posted thisA 60% reduction in token usage for coding agents with one CLI? That is the kind of sensational claim built to spread. So I tested it on a real codebase. The claim was that ast-grep could make coding agents significantly more effective, or use roughly 60% fewer tokens. That sounded plausible. AST search can return structured results instead of large blocks of text. So I installed it and tested it against the whole my-pi codebase: • 964 source files • About 287,000 lines • TypeScript, JavaScript, and Svelte • Compared with ripgrep, LSP tools, and Fallow The result was more nuanced. ast-grep was useful for targeted structural searches and code transformations. But it did not improve general repository understanding enough to justify making it part of my agent workflow. For understanding files, LSP document symbols provided richer information. For broad discovery, ripgrep was simpler and faster. The problem I care about most is agents creating local code that duplicates an existing utility. Fallow found 22 clone groups and 673 duplicated lines automatically. ast-grep needed me to know the duplicated syntax before I could search for it. Its patterns could also be fragile. TypeScript return annotations caused valid functions to be missed, and its outline command returned nothing for the Svelte files tested. So I removed it. The lesson was not that ast-grep is bad. It is a strong tool for structural lint rules, migrations, and controlled rewrites. But adding another tool does not automatically give an agent better understanding. Good agent tooling needs clear roles: • rg for text discovery • LSP for symbols, types, and references • Dedicated analysis for duplication • Searchable storage for large outputs • Guardrails that enforce project conventions That is the approach I am taking with my-pi: a curated Pi coding-agent distribution with LSP, MCP, skills, recall, guardrails, redaction, telemetry, observability, and a context sidecar already wired together. I would rather test a tool on real work and remove it than keep it because a benchmark sounds impressive.
-
Scott Spence liked thisScott Spence liked thisWhat happens in Vegas doesn’t always stay in Vegas… Our colleague Nemo recently attended DEF CON, one of the world’s largest cybersecurity conferences. He put together this short video sharing his experience, what caught his attention and what he brought back to the Cloud Lobsters team. Take a look 👇 Thanks to Claire Whitaker at Sequentially for the video production. Scott Meechan Tara Whitaker Duncan Elliot Nemo Godebski-Pedersen Scott Spence Yehen Singankutta Arachchilage #DEFCON #Cybersecurity #CloudLobsters #Insurance #insurtech #artificialintelligence
-
Scott Spence liked thisScott Spence liked thisMeet Yehen Singankutta Arachchilage, a third-year Computer Science student at Imperial College London who’s been putting his skills into practice as a Software Engineering Intern at Cloud Lobsters. Working at an insurtech startup has given Yehen the opportunity to contribute to real products, tackle real engineering challenges, and see first-hand what it takes to build systems in a fast-moving startup environment. For students like Yehen, internships at early-stage companies can offer something particularly valuable: the chance to take ownership, learn quickly, and make a genuine impact. Hear from Yehen about his experience at Cloud Lobsters and what he’s learned along the way! 🦞 Thanks to Claire Whitaker at Sequentially for the video production. Scott Meechan Tara Whitaker Duncan Elliot Nemo Godebski-Pedersen Scott Spence #insurtech #insurance #digitaltransformation #artificialintelligence
-
Scott Spence reacted on thisOne thing I have learned over the years is that the best engineers rarely need convincing. They want interesting problems, ownership and talented people to work alongside. That's exactly the culture we are building at Cloud Lobsters. If Tara's post resonates with you, we would love to hear from you.Scott Spence reacted on thisOur strongest engineering hires usually found us first. So instead of asking where to find our next engineer, we asked ourselves a different question: What makes Cloud Lobsters the sort of place engineers enjoy working? We're an InsurTech company at heart, and that's where we've built deep expertise. Insurance is full of complex data, workflow and automation problems. But good engineering isn't tied to a single industry. Last year alone our engineers worked on AI-powered underwriting, banking workflows, logistics software and industrial optimisation. The domain changes. The engineering challenge is what keeps it interesting. That's reflected in how we build software: • 𝗣𝗿𝗼𝗯𝗹𝗲𝗺 𝗳𝗶𝗿𝘀𝘁, 𝗰𝗼𝗱𝗲 𝘀𝗲𝗰𝗼𝗻𝗱: We work directly with clients from day one. We spend time understanding the business problem, challenging assumptions, prototyping quickly and then building the right solution. • 𝗣𝗿𝗮𝗴𝗺𝗮𝘁𝗶𝗰 𝘁𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆: We use Svelte extensively, embrace AI where it genuinely adds value and choose technology because it solves the problem—not because it's fashionable. • 𝗕𝘂𝗶𝗹𝗱 𝘄𝗲𝗹𝗹, 𝗿𝗲𝘂𝘀𝗲 𝗼𝗳𝘁𝗲𝗻: When we create something genuinely reusable, it becomes part of our toolkit. Every project leaves us with better tools, so we spend more time solving new problems instead of rewriting old ones. • 𝗙𝗮𝘀𝘁-𝗺𝗼𝘃𝗶𝗻𝗴 𝗽𝗿𝗼𝗷𝗲𝗰𝘁𝘀: Most engagements last weeks or months rather than years, so you'll work across different industries, solve a wide variety of problems and see the impact of what you build. We keep our teams small, experienced and collaborative. You'll work alongside engineers who challenge each other, share what they know and care about building great software. Everyone contributes, everyone learns and everyone ships. We're growing our engineering team in London. If you enjoy owning problems from the first client conversation through to production, you'll probably feel at home here. You'll probably enjoy working with us if you: • enjoy understanding the business before writing the first line of code • care about writing clean, maintainable software • enjoy ownership, direct client interaction and project variety • want to work together in our London office If that sounds like you, drop me a message. I'd love to have a conversation.
Experience
Education
Publications
-
Moving from beginner to (slightly more) advanced git with aliases.
Medium Hacker Noon
See publicationGet familiar with git aliases and speed up your git workflow
-
How to build and deploy a multifunctional Twitter bot
Medium freeCodeCamp
See publicationBuild a bot, tweet a picture, make a markov bot, tweet links from a spreadsheet
-
Why you should have your own Twitter bot, and how to build one in less than 30 minutes
Medium freeCodeCamp
See publicationLightly edited by the publication from my original title of "Make your own Twitter bot in under 30 minutes"
If you need a Twitter bot to retweet and like Tweets on specified hashtags and keywords then you can have one setup in under half an hour
Projects
-
SkyKit
- Present
Free analytics and statistics for Bluesky users. Track post engagement, find inactive accounts, and get insights with no authentication required. Built with SvelteKit. https://skykit.blue
-
Svortie
- Present
AI-powered SvelteKit application platform with Daytona sandbox integration. Runs autonomous coding agents in isolated environments for safe, parallel development workflows.
-
DevHub CRM
- Present
Developer-first CRM for managing relationships through GitHub, conferences, and open source. Features GitHub Quick Connect, relationship health scoring, and QR codes for conference networking. Built with SvelteKit, Svelte 5, and SQLite. https://devhub.party
-
Sveltest
- Present
Comprehensive testing resource for the Svelte ecosystem. Includes AI assistant rules for Cursor and Windsurf, battle-tested patterns from production environments. Built to generate coding assistant rules that improved team productivity. https://sveltest.dev
-
MCP Tools Suite (20+ tools, 1,200+ GitHub stars)
- Present
Suite of 20+ Model Context Protocol (MCP) tools for enhancing AI coding assistants. Covers search, memory, workflow automation, and documentation access. 1,200+ combined GitHub stars. Top tools include mcp-sequentialthinking (553+ stars), mcp-omnisearch, and mcp-memory.
-
Svead
- Present
SEO metadata component for SvelteKit. Used in production by Ashley Furniture. Handles meta tags, Open Graph, Twitter cards, and JSON-LD structured data.
-
SvelteKit Embed
Embed third party media in your Svelte projects
Languages
-
English
-
Recommendations received
1 person has recommended Scott
Join now to viewOther similar profiles
Explore more posts
-
James Macauley
Software Engineer CVs vary wildly - some are so much more impressive then others. Aside from the obvious stuff like - mentioning the language you used - (yes, this happens) The difference between an great CV which gets you interviews, and a poor CV which gets rejected - is maybe only a couple of key things. 📈 - Engineers often only state what they used, or what they did. But they are missing the interesting part of that, which is impact and outcome. What was the result to the business and to the customer? - was it faster, more reliable, did it solve a historical problem, did it save money? 💭 - How did you do it? And why did you do it that way? Going down a level and discussing the architecture and approach is so fundamentally important, but yet it is surprisingly so rare. If it is a microservices, event driven architecture you need to mention that. It is important and will get you interviews. 🔎 Do not make assumptions. Add detail. You should include details such as the industry sector, the dates including the months of your employment, the technology outside of the language you worked on - such as Sidekiq, Express, Next - things you might omit. Many engineers will assume that something is obvious. You’re giving too much credit to the person who might be reading it. Yes CVs vary wildly, but so does the opinion of someone reading your CV. You're better to make something clear, then assume someone will read between the lines. - Software Engineers underestimate the importance of their CVs. There are unhelpful pre-conceived ideas, but as someone who looks at CVs literally all day long - the engineers who take the advice, and invest the time to get it right are the ones getting the interviews and the best jobs.
56 Comments -
Coburg Banks IT Recruitment
🚨 **The Hidden Cost of Hiring the Wrong Developers** 🚨 Think a bad hire is just a small misstep? Think again. When it comes to developers, the impact can hit harder than you realise—delayed projects, wasted budgets, low morale, and slipping deadlines are just the start. Discover the *real* cost of hiring the wrong IT talent and learn how to avoid these costly mistakes. If you’re tired of recruitment headaches and want to build a stronger, more efficient tech team, this one’s for you! Your next hire could make or break your business—don’t leave it to chance. 👨💻 👩💻 #ITRecruitment #TechTalent #HiringMistakes #CoburgBanks #RecruitmentSolutions
-
FUNCODE SRL
Kotlin Property Delegates You’re Probably Not Using: observable and vetoable Why Property Delegates Often Go Unnoticed Most Kotlin developers know lazy — it's everywhere. You see it in Android ViewModels, dependency injection setups, and library code. But the standard library ships two more property delegates that are surprisingly powerful and almost never discussed: Delegates.observable and Delegates.vetoable. If you've been wiring up manual setters or using LiveData just to observe a simple field change, these might be exactly what you've been missing....
-
Wave Group
London-based JavaScript engineers, don’t sleep on this👇 The JavaScript London meetup is quickly becoming a go-to space for engineers who want more than just another event. It’s a mix of learning, conversation, and genuinely meeting people in the industry, with a strong group of regulars starting to form. If you’re looking to stay sharp, connect with others, and hear from interesting voices in the space, keep an eye out for what’s coming next. #JavaScript #LondonTech #Meetups #Engineering #TechEvents
-
Pocket
If you build software for clients, listen up... Most freelance and contract developers have insurance. Far fewer have checked whether it covers the thing they're most likely to be claimed for because a standard Professional Indemnity policy, the common default, often excludes technology-specific failures. What a developer actually gets claimed for is rarely dramatic: → a bug that reaches production and costs the client money → a build that doesn't do what was contracted → lost or corrupted client data → a project that runs late and causes a loss None of these requires negligence in any obvious sense. And the claim isn't capped at your project fee... a bug in a small job can cause a loss many times its value. The cover that responds is IT Liability (ICT Insurance): Professional Indemnity built for technology work, paired with Public Liability, with IP protection usually sitting inside the PI section. The tech-specific version matters here... a generic PI policy can leave you exposed for your most likely claim. Easy-to-understand guide: https://lnkd.in/gKWgvUNc General information only and not personal advice.
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore More