Joe Mathews · CTO · thinker · builder · category creator
CTO at Duel. I build for the modern era without rewriting the world. Money and data stay slow and sacrosanct. The product layer goes fast and experimental. Ship new products in days. Your agents never sleep. The audit trail outlasts the founder.
That's not tacking agents onto a 2018 Jira-and-Scrum workflow. It's Conway's Law as a design tool. The org chart and the architecture are the same diagram. All backed by Context Capsules, Bimodal architecture, and the Engineer → Builder shift to conscious orchestration.
Twenty-five years at the forefront of every shift that mattered: Telecoms, Mobile, online video, IPTV, FAST, ad-tech, ML, AI, fintech, ed-tech, LLMs, and now brand advocacy. The structural debt is always the same. The build needed is always different.
Built tech at
Duel · GoStudent · CogX · FoodHak · Amplyfi · Sorenson (Nielsen) · Xumo (Comcast) · MySpace · Vodafone · Zenulta
Founded
LeaveNoPrints · Gower Electric Bikes
Creating the brand advocacy category with Duel whilst taking the existing platform and team AI-native. Bimodal platform + Engineer → Builder + Context Capsules adopted in production across engineering, AI, and data. The team and the platform changing shape together, on purpose.
Context Capsules: operational context for human and AI agents. Public spec, CC BY 4.0, adopted in production, presented at AWS Summit London 2026 (4.83 CSAT, second only to a session with only 4 votes). contextcapsules.com
More talks being lined up. Regular CTO round-tables; chair “The Few”, active in others. Always open to more for inspiration and sharing my wealth of experience.
A small number of founders and CTOs. Advisory engagements have led to acquisitions. I'm careful about how many I take on at a time.
favmeup, mydreamhandle, saasyearend weekend vibes. EdTech, productivity tooling, and numerous personal skills used daily, not publicly released.
How the pattern works
Money and data move slow and considered: governed, audited, never breaking. Product moves fast and experimental: UI in days, disposable where it needs to be. The team structure mirrors the split. Conway’s Law working for you, not against you.
Sacrosanct · slow
Considered cadence. The platform layer that handles money, identity, integrity, and the event backbone. Built once, well. Outlasts the founder.
Experimental · fast
Experimental cadence. Surfaces, features, AI workflows. Vibe-coded prototypes harden into patterns. What learns is folded back into the platform; the next team gets it for free.
The same SDK serves both. The same team structure mirrors both. The same event backbone connects both.
Opinions
These are the named principles I keep coming back to, in the work, in the writing, on stage. Pick any one and the rest follow. Steal them; they travel.
Bimodal architecture
Money and data move slow and sacrosanct. Product moves fast and experimental. Two cadences, one event backbone. The org chart is the architecture diagram. Change the team, change the system, on purpose. This is what makes Bimodal actually work in an AI world: governance baked into the structure, not bolted on after.
Context Capsules
Not docs. Not a wiki. Not a Claude Skill or a Cursor Rule. A structured operational briefing covering What, Bounds, Gotchas, How, so anyone, human or AI, can act on a concept correctly without a 45-minute briefing from the one person who knows. Authored, public, CC BY 4.0. Already in production at scale. contextcapsules.com
Engineer → Builder
Engineers move to orchestration. They take an end-to-end problem, decide what an AI should solve and what needs deeper human judgement, and ship the answer. The unit of work is the outcome, not the ticket. You can't build a team today the way you did two years ago, and the people who pretend you can are the bottleneck.
Vybrid coding
Production-grade software in the AI age doesn't come from one mode. Vibe-coding proves the experiment in a day. The discipline is folding what works into the platform so the next team gets it for free. Speed without scar tissue, scar tissue without speed loss.
Bloody Brilliant Builders
Proactive. Customer-obsessed. Hunt for value. Orchestrate human and AI work. Ship. The team isn't shaped like it used to be, it can't be, and the people aren't either. Outsourcing is yet to catch up. By the time we've defined what we need, we've built it.
Revenue per FTE engineer
When AI compresses how much one builder can produce, headcount stops being the right denominator. Revenue per full-time engineer is the number that tells you whether the shift is actually working, and whether the team is structured for the era it's in.
Make work visible
Cadences fail when only feature delivery is treated as work. The cadence I run gives equal space to discovery, design, prototyping, technical debt, testing, and BAU. It makes the trade-offs legible, and stops the urgent from always beating the important.
Resumability as a kindness
The next person, your teammate, your AI agent, your future self, should be able to pick up the work without a re-briefing. Bound the context. Document what isn't included. The kindness is structural; the productivity gain is the side effect.
Engineering Hubs
A small number of offices in tech-rich cities, with employees hired inside a commute radius. You get the talent pool of remote with the planning velocity of in-person. Follow-the-sun support emerges naturally. Local cultures form, feed innovation, and stop one location's myopia from running the whole org.
Solid engineering practices
Service Orientated Architecture. Clear Context and Decisioning. Up-to-Date Documentation. Architectural Patterns. Design Systems. Non Esoteric Code. Don't be too DRY. All signals of solid engineering, that are now even more prescient in the AI age.
Talks & writing
Structured operational context for humans and AI. The Minimum Viable Context you need. Nothing more.
Talk on Context Capsules. 4.83 CSAT, second only to a session with only 4 votes.
What replaces SEO and influencer when the front door to a purchase becomes an AI assistant. Verified human advocacy as the signal AI systems can't get anywhere else.
Where I think out loud, mostly in the comments. Lifelong lurker.
The long arc
I've been through the shifts. Categories created, dead launches at scale, AI before the LLM wave, teams rebuilt. Pre-seed to Series E. The interesting thing isn't the dates; it's what kept working.
Creating the Brand Advocacy category at Duel. Bimodal AI-native platform, Context Capsules adopted across engineering + data, non-AI team taken to AI-native.
Senior Director of Engineering at GoStudent. 70+ engineers, attrition 30% → 4.5%, eNPS −34 → +50, led technical diligence on the £95m debt round at a $3.4 billion valuation, MarTech transformed from task-driven bottleneck to modern Product Engineering.
Product Strategy / Head of Engineering at CogX. Walked in, defined the product vision, pitched it to private investors, walked out with a 6-figure cheque inside 40 minutes. Then built the team to deliver, working with some of the people who built the founding technologies the rest of us use every day. Released v1 of the CogX Insights App to 300% MoM user growth.
I've tried, I've failed, and once I succeeded in the wrong way. Founded LeaveNoPrints (Shop Online, Plant Trees). Integrated with the top European affiliate networks; 1M+ product catalogue. Quietly parked. Re-releasing for the current climate. Co-founded Gower Electric Bikes, online e-bike rental on a UK Area of Outstanding Natural Beauty. Welsh-Govt funded, Green Goals gold award, 5-star reviews, top SEO. Too successful in the wrong way: too physical to scale beyond the Gower, never enough to replace the day job. Ran as a lifestyle business into its fourth season.
Founding CTO at FoodHak. Stood up the in-house AI team and used ML to prove the link between Amla and cholesterol reduction greater than $5/day commercial statins. Foundation for scientifically validating Ayurvedic ingredients.
EVP Product, Engineering at Amplyfi (Wales' first unicorn-tagged AI startup). Built the technology org from 6 to 50+. Led the £5m raise. Cut AWS by ~£500k/year. Scraping the whole web for trend prediction, years before LLMs made AI mainstream. Published research with CSET and UK/US universities.
VP Product, Engineering at Attrix (Sorenson Foundry, sold to Nielsen). Stood up Engineering, QA, DevOps, and Product from scratch. Designed and filed the patent for a system that recouped up to 50% of the parent company's lost ad inventory, its primary revenue source.
VP Engineering at Xumo (acquired by Comcast for $100m+, now the core of their IPTV offering). Founding engineer, 0→50. Created the FAST (Free Ad-Supported Television) category. Ran the team to launch LG Channel plus, now available globally. Dead Drop product launch: 0 to 60M users on deploy, first user in under one second. Founding member of the Smart TV Alliance. VAST/VPAID v4.0 spec advisor. Patents filed for high-resolution UI on low-end CE devices.
Telecoms at Vodafone and Zenulta. Contract negotiations in the morning, code in the afternoon. Multiple-Quick-Wins-in-7-days kind of projects. The cadence stuck. Years later it's the same shape: ship the smallest thing that proves the bet, harden what works, never lose the thread between the commercial reality and the code.
The full reverse-chrono is on LinkedIn . If you want a CV, ask.
Where I'm sharpest
Advisory, fractional, occasionally full-time-for-the-right-thing. Each of these is something I've shipped, not something I read about. Pick the one that sounds like your problem.
I've done it from the inside. Restructure where it's needed. The team transitions, the operating model, the cadence that makes the trade-offs visible, the hire bar, the orchestration playbook.
Strangle the monolith without rewriting the world. Bimodal architecture, sacred slow plus fast experimental, Context Capsules as the bounded-context layer. Money and data stay safe. Everything else moves insanely fast.
Engineer → Builder. Production-grade vibe coding (Vybrid). Reusing the engineering patterns that survived every previous shift. Revenue per FTE engineer as the number you actually track.
I've shipped at every stage. The shape of an engineering org that wins in the AI age is not the shape that worked five years ago. I'll tell you what to keep, what to drop, and where the bodies are buried.
Context Capsules across the business, with Finance using the same definitions as the AI building the engineering dashboards. The operational source of truth that makes everything else faster.
Regardless of your industry, I'm wired to get in and understand where your company stands in a matter of days and weeks. Providing a fresh perspective, I've repeatedly realigned both product and tech visions to provide exponential growth opportunities.
Bounds: what this isn't
I take on a small number of advisory and CTO-level engagements at a time. I don't run a body-shop. I don't pretend to be your full-time CTO from a beach. If you need a hands-on CTO indefinitely, hire one. If you need a sharp outside view, a steady hand on the AI-native transition, or someone to be the face of tech with your board for a stretch, there's likely none better.
From people I've worked with
“If you are on the market for an experienced tech leader who inspires your org and rolls up his sleeves to help his team deliver, look no further”
“His expertise in technical leadership coupled with his ability to support scaling both technology and operations made him an invaluable asset”
More on LinkedIn .
Get in touch
No form. No Calendly. Email is fine. So is LinkedIn.