About
Most CTOs are good at one thing. Strategy or execution. Vision or code. I've spent 15…
Services
Activity
2K followers
-
Mathew Dostal shared thisLeverage compounds; effort just accumulates One morning, I woke up to 124 open pull requests. About 90 were duplicates. My AI coding swarm had spent the night doing the same tickets over and over because I never told it to check whether a PR already existed first. That was the good news. The work happened while I slept. And the failure was a system I could fix once, instead of a mistake I'd have made by hand a hundred times. The swarm runs 35 concurrent sessions across 6 repos on two machines in my house. 6,500+ runs in 7 weeks. But the swarm isn't the point. It's the most extreme version of one idea: leverage compounds; effort just accumulates. You are the bottleneck you can't see. The work that only routes through you is the work capping your whole team. It feels important. It's a ceiling. And leverage is built, not found. Hire the person who replaces you and actually hand them the keys. Document the thing in your head so the team stops needing you in the room. Buy the managed service instead of babysitting the one you run. Here's the part founders resist: indispensability is a bug. If the work stops the moment you step away, you're not valuable. You're the critical path. I wrote up the whole system: the 5-layer memory, the 3-layer architecture, and every way it broke. Link in the comments. Here's the thing: this setup is already aging. Seven weeks in production and the ground's shifted under it — I'm already building the next version. That's the actual job. Not the swarm. The willingness to obsolete your own best system before it obsoletes you. If you're running agents, what broke first for you? For me, it was memory: they forget everything the moment a session ends. #AI #Agents #TechLeadership #FractionalCTO
-
Mathew Dostal shared thisCulture is behavior under pressure, not values on a wall The values on your wall are what leadership wishes were true. The behavior under a deadline is what's actually true. When the two disagree, the wall loses. And your best engineers already know it. The 6,000% rebuild I did at Hertz didn't come from me. It came from a team that trusted each other enough to rearchitect a $10B pricing engine. Jerry invented the hashing strategy that made the cache work. Aiden built the testing infra that proved it was stable under load. When I wrote that story up, I named all of them. Not as a courtesy. As the point. How you handle credit under pressure is culture. Who you protect when something breaks is culture. The pizza party is not. Two failure modes show up everywhere, and both wear a cape. Hero culture, where the org quietly depends on one person grinding through every crisis and calls it strength: that's a single point of failure you decided to pay a salary. And knowledge hoarding, where information becomes job security and your smartest person becomes your worst bottleneck. I've made the burnout version of this myself. The cost didn't show up in the sprint. It showed up three months later, in the resignations. How you treat the people who built the thing isn't an HR footnote. It's throughput. Full write-up in the comments. What's the gap between your values on the wall and what actually happens at 9 pm on a Friday before a launch? #EngineeringLeadership #Culture #CTO #Hiring
-
Mathew Dostal shared thisThe expensive failures are decisions, not bugs. Any strong senior engineer can pick your database for $150 an hour. That's the cheap part of the job. The failures that actually cost you aren't bugs — they're decisions. And the most expensive ones aren't about what to build. They're about how to get there without the wheels coming off mid-flight. Chick-fil-A, $20M EY engagement: replace the POS across 2,600+ locations. The new system had to assume the network drops and keep taking orders. But the old system was already live in thousands of stores — partially offline-capable itself — and it couldn't go down for a single lunch rush. You don't get to flip a switch. The whole risk lived in one question nobody owned: how do you migrate something that big, that is live, without a moment where it's broken? That was the piece I owned — the sequencing. The iterative milestones. Where each cutover happened, what got phased out when, what the rollback was at every step, and which pieces had to stay dual-running until the new path proved itself under a real lunch rush. The business couldn't specify that. Engineering couldn't authorize it alone. It lived in the gap between them — and that gap is the actual job. A big-bang cutover would've demoed beautifully and taken down a thousand stores. The unglamorous call — roll it in pieces, prove each one, never bet the whole chain on a deploy — is what made it survivable. Title and tenure don't tell you who makes that call. I've met people with a decade of seniority who'd have flipped the switch. Judgment shows in the restraint. So when you're sizing up technical leadership, don't ask whether they can build. Most people can build. Ask how they'd roll it out — and what they'd refuse to do all at once. I wrote the full breakdown of what a CTO actually does all day. Hint: it's not picking databases. Link in the comments. What's a rollout you've seen go sideways because someone bet the whole thing on one cutover? #CTO #FractionalCTO #TechLeadership #Architecture
-
Mathew Dostal shared thisRates at Hertz changed 10,000 times per second. We didn't cache them. That sounds wrong. It was the decision that made everything work. The expensive part of a rate lookup was never the rate. It was figuring out which of tens of thousands of rules applied to that one request. That ran on every call, hundreds of DB hits deep. So we cached the rules, which barely changed, and left the rates alone. Hundreds of round-trip flights became two. The worst case dropped from about 30 seconds to 30 milliseconds — and throughput climbed roughly 6,000% on a quarter of the infrastructure. The lesson wasn't "cache more." It was "cache the right thing." Most of the complexity I find inside companies is the opposite of that. Kubernetes is running for 20 people. Terraform for a single box. A custom-built cache, a set of complex microservices that deploy together, or an internally managed service that could have been handled for low cost or free. It's not engineering. It's anxiety with a build step. And it bills you every month: maintenance, one more thing to break, the one person who understands it, and can't be replaced. The senior skill is subtraction. Knowing what not to build is harder than knowing how, because subtraction never photographs well for the architecture diagram. I wrote up the full breakdown: where complexity earns its keep, where it's just fear, and how to tell the difference. Link in the comments. What's the most over-built system you've inherited? #Architecture #TechLeadership #CTO #FractionalCTO #DevOps
-
Mathew Dostal shared thisGoogle studied tens of thousands of their own interviews and found zero correlation between interview scores and job performance. Zero. The company that invented the modern technical interview proved it doesn't work. And the industry kept doing it anyway. Here's what's worse: researchers found that candidates in private coding interviews performed twice as well as those being observed on a whiteboard. All women in the public interview condition failed. All women in the private condition passed. The interview wasn't measuring ability. It was measuring anxiety tolerance under performative stress. I've hired dozens of engineers across seven companies. The ones who turned out to be great almost never stood out in a whiteboard round. They stood out when they walked me through something they built or in pair programming where I could see how they actually think. They stood out in a contract where the work spoke for itself. The work IS the interview. Everything else is theater we've convinced ourselves is science. I wrote the full breakdown ; what actually predicts success and how to structure interviews. Link in the comments. #Hiring #EngineeringLeadership #TechHiring #CTO #FractionalCTO #SoftwareDevelopment #TechnicalInterviews
-
Mathew Dostal shared thisI spent years at EY and Zilker running full technical engagements: 40+ hour weeks where architecture decisions were buried under backlog grooming and PR reviews. The high-value work: direction and the "why" behind the tech: took maybe 10-15 hours a week. The rest was overhead. That’s the difference with Fractional: it strips away the noise. I work with companies at every stage now: - Enterprise teams needing architecture direction for specific initiatives. - SaaS startups needing critical early decisions from someone who's done it before. - Non-technical founders needing tech, growth, and marketing strategy unified. The pattern is clear: most companies don't need 2,080 hours of a CTO annually. They need 500. Maybe 200. Maybe 40. My new post breaks down the math, the engagement stages, and when you actually should go full-time. Read more here: https://lnkd.in/grg_-8Tq #FractionalCTO #StartupLeadership #TechnicalStrategy #CTO #Entrepreneurship
-
Mathew Dostal shared thisRates at Hertz changed 10,000 times per second. We didn't cache them. That sounds wrong, but it was the decision that made everything work. The expensive part of a rate lookup wasn't the rate : it was figuring out which of tens of thousands of rules, discounts, and eligibility criteria applied to that specific request. That filtering was happening on every single call, hundreds of DB hits per request. We moved all of that work upfront. Pre-filtered rules grouped by location and corporate account into Redis. When a request came in: one cache fetch for the rules, one targeted DB call for the rate. Hundreds of round trips became two. The result: p95 latency went from ~3 minutes to 30 milliseconds. 6,000% improvement. And we did it with a quarter of the infrastructure : 10 Node instances instead of 40. I wrote the full architecture breakdown : CDC streaming, Kinesis evolution, geo-routed replication, and the team that made it happen. https://lnkd.in/gZs-ZvxK What's the most counterintuitive caching decision you've made on a system? #architecture #techleadership #softwareengineering #performance #redis #fullstack #devops
-
Mathew Dostal shared thisI just published one of the biggest tech glow ups I’ve worked on: the architecture behind a 6,000% performance improvement at Hertz’s rate engine. But I’m most proud of the team mentioned at the end. Jerry (Gerardo Leon) invented the hashing strategy for the cache. Taffy led the charge through the complex dev: and added SLAYER as a test code. Aiden built the testing infra to prove stability under load. Two Aarons, Don, Andrew, Dominika, Layne, and the original foundation team. Plus Tim, our first lead who set the stage for our growth. I keep forgetting people, so feel free to comment or call me out. I should have given credit to our awesome project managers: Todd and then Leigh Hunter. Leigh was there for the growth from the initial release through the massive performance improvements and had to keep up with trying to drive that as the contracts were not filled until we met the much higher SLAs. I'm sure I've forgotten more people and if you want me to tag you instead of casually mentioning or you want a shout out, let me know! I even rewrote the e2e test suite from JMeter to Taurus on a flight to France: slow tests are tests nobody maintains. The deep dive covers CQRS, CDC streaming, and Kinesis, but the real story is what happens when a good team trusts each other to rearchitect a $10B company’s pricing engine. Full story: https://lnkd.in/gZs-ZvxK #SoftwareArchitecture #CloudComputing #TechLeadership #Innovation #FullStack #Hertz #PerformanceOptimization
-
Mathew Dostal shared thisIn 2015, I drew a CI/CD pipeline diagram at Kohl's. Build, Test, Publish, QA, Deploy, Stabilization, Release. That loop required hundreds of people : developers, QA engineers, DevOps, project managers, release coordinators, and a branching strategy document that went through 11 revisions. I found that diagram last week. It maps almost 1:1 to the ticket lifecycle of the autonomous coding swarm I built on two machines in my house. Same stages. Same gates. Different executors. The swarm runs 35 concurrent AI coding sessions across 6 repos. It scans the ticket board every 2 minutes, creates isolated git worktrees, spawns Claude Code sessions, writes code, creates PRs, and updates ticket statuses. I wake up to pull requests. 6,500+ agent runs over 7 weeks. A 5-layer memory architecture so agents don't repeat each other's mistakes. And a catalogue of failures that taught me more than the architecture itself : like the morning I woke up to 124 duplicate pull requests because the director never checked if a PR already existed. I wrote the full technical breakdown: the 3-layer architecture, the memory system, the code that runs it, and every way it broke. I wrote more about this at https://lnkd.in/gf82zRd7 If you're building autonomous dev tooling or thinking about it : what's the biggest gap you've hit? For me it was memory. Agents forget everything the moment a session ends. Solving that changed everything.
Mathew Dostal replied to a comment
1d
Moonveil AI, review capacity, easily. Generating PRs is cheap now — reviewing them at that volume is where I actually became the bottleneck. Still testing the full solution and I will post when its ready.
Mathew Dostal replied to a comment
1d
Thanks Benjamin Torres — the memory piece is what I'm most curious about across other setups. Curious what's broken first for you if you're running agents.
-
Mathew Dostal liked thisMathew Dostal liked thisI came across this clip on YouTube today. What caught my attention wasn’t the pitch—but how calmly and confidently a young child handled rejection, pressure, and negotiation in real time. 𝐊𝐞𝐲 𝐎𝐛𝐬𝐞𝐫𝐯𝐚𝐭𝐢𝐨𝐧𝐬 - When investors said “I’m out”, he acknowledged it instantly: “I appreciate it.” - When an offer came, he quickly calculated the numbers, assessed the value, and confidently declined—without hesitation or emotion. - No panic. No over-explaining. Just clarity and presence of mind. 𝐒𝐨 𝐰𝐡𝐚𝐭? This is a powerful reminder of what leadership looks like under pressure: - Staying composed when options shrink - Thinking on your feet with limited time - Respectfully saying no when the deal isn’t right Real confidence isn’t about speaking louder. It’s about clarity, composure, and conviction. 𝘝𝘪𝘥𝘦𝘰 𝘴𝘰𝘶𝘳𝘤𝘦: 𝘠𝘰𝘶𝘛𝘶𝘣𝘦 #Leadership #DecisionMaking #Confidence
-
Mathew Dostal reacted on thisMathew Dostal reacted on thisThe Best $200 Ever Spent? 💰🔥 Back in the early days, Nvidia’s Jensen Huang was asked how much cash he had in his pocket to help incorporate the company and price the shares... 🧐 The answer? $200. 💵💸 That "pocket change" bought him roughly 20% of the company. 📈 Today, that same stake is worth hundreds of billions! 🤯💎 Talk about a legendary return on investment! 🏦✨ Sometimes, the smallest beginnings lead to the biggest revolutions! 🌍🚀 #Nvidia 🟢 #JensenHuang 👨🏻🍳 #TechHistory 📚 #Entrepreneurship 💡 #Investing 📊 #StockMarket 📉 #StartupLife 🚀 #SuccessMindset 🧠 #AI 🤖 #GPU 🎮 #WealthBuilding 💰 #Fortune 💸
-
Mathew Dostal liked thisMathew Dostal liked this99% of people fail the "Sell me this pen" test. Here is why. We’ve all seen the movie. We’ve all heard the line. But Jordan Belfort (the real Wolf of Wall Street) says almost everyone gets the answer wrong. If I hand you a pen and say "sell me this," what do you do? ❌ The Amateur: Starts talking about the pen. "It’s a great pen, it writes upside down, it’s cheap." ✅ The Pro: Starts asking questions. Belfort explains that sales isn't about ramming a product down someone's throat; it's about qualification. When you immediately start pitching features, you are saying, "I don't care about you, I just want to make a sale." Instead, the only rational response is to ask: "How long have you been in the market for a pen?" "What kind of pens do you usually use?" "Is this for business or personal use?" The golden rule: If they tell you they aren't looking for a pen, you don't try to sell them one. You shake their hand and wish them a nice day. Rookies try to sell to everyone. Masters only sell to those who have a need. Stop pitching. Start interviewing #marketing #branding #sales #entrepreneurship #management
-
Mathew Dostal liked thisMathew Dostal liked thisShe was offered $95K. She countered with $180K. Everyone's mocking this Gen Z candidate. She's not the problem. Here's what actually happened → Virtual interview. Salary question. She says money isn't her focus, then the range drops: $70K to $95K. Her counter: $180K. Her logic: "I'm solving six figure problems. That range is for average people. I'm not average." Interview ends with "Okay whatever, I'll talk to you later." The uncomfortable truth → She's not wrong about the value she creates. She's wrong about thinking anyone cares. We spent decades selling meritocracy. Work hard, get rewarded. Build skills, earn your worth. Then we paid people based on "market rates" and "budget constraints." Translation: we pay what we can get away with, not what you're worth. Gen Z watched their parents stay loyal to companies that laid them off in PowerPoint slides. Watched high performers get 3% raises while mediocre executives got 40% bonuses. They saw the game. So they stopped playing by our rules. What everyone's getting wrong → The real controversy isn't her ask. It's that we're shocked she asked. We built an entire system on labor negotiating from gratitude. Show up humble. Accept the range. Prove yourself first, get paid later. That social contract is dead. Not because Gen Z killed it. We did. Every time we posted record profits while freezing salaries. Every time we paid consultants $500 an hour but told employees $75K was "competitive." They're not entitled. They're pattern matching. The part that makes us uncomfortable → She might actually be right. Not about deserving $180K based on potential. But about refusing to accept manufactured scarcity in compensation discussions. Most salary ranges are fiction. Anchored to last year's budget, not this year's value. Designed to lowball candidates who don't negotiate. When she said "that range is for average people," she wasn't being arrogant. She was naming the game we pretend doesn't exist. Here's the conclusion no one wants → This isn't about one bad interview. It's a stress test of every assumption we hold about compensation and leverage. And we're failing it. Not because she was right to ask for $180K. But because our only response is mockery, not reflection. We're laughing at her audacity instead of asking why an entire generation thinks this is reasonable. The generation entering the workforce has fundamentally different assumptions about loyalty, meritocracy, and fair exchange. We can adapt our thinking or keep writing posts about how kids these days don't get it. One is strategy. The other is just nostalgia dressed up as wisdom. VC-Internet
Experience
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Education
-
Neumont College of Computer Science
-
-
Activities and Societies: Student Government Senator, GC Order Senator
Completed a 4-year Computer Science degree in 2.5 years through Neumont's accelerated, project-based curriculum.
-
-
Volunteer Experience
Honors & Awards
-
Rising Star
Ascendant Technology
Award for a recently joined, high performing individual.
Languages
-
English
-
Recommendations received
1 person has recommended Mathew
Join now to viewOther similar profiles
Explore more posts
-
Paul Chiusano
I thought this was a good and very reasonable talk by Eleanor Millman on how to prioritize projects for a platform engineering team: https://lnkd.in/gaYtUHRH A few insights I took from it: - Unlike product features which can be more directly tied to revenue, platform engineering work is a couple steps removed. But that doesn't mean it's unimportant! Far from it, platform engineering projects can make everyone at the company more efficient, able to produce higher quality software, etc. - While "urgency" is always going to be a factor, you don't want to just be putting out fires, you want be able to prioritize work which is "high impact" ... even if that impact isn't felt immediately. - Good rule of thumb: prioritize "highest impact for lowest effort". - The talk has some ideas on what factors to choose for impact, and how to blend them. For instance "speed of development" is one factor, "cloud cost optimization" might be another. The weighting of different impact factors can change over time, depending on the needs of the business. I would say that a lot of companies don't have much methodology here but I can really see the value in codifying it. You can always change the methodology or the weighting if it's spitting out results that don't pass the smell test or you really feel it is leading the company astray. Clarity can give the org more freedom to put resources behind projects that would otherwise never be taken on. When things are unclear, prioritization still happens implicitly, but the decisions tend to be a lot more random and fear-based and the org is worse off as a result.
3 Comments -
Dexter Cousins
A CTO commented on my post last week. AI Productivity failures aren't a leadership problem, he said. They're a legacy technology problem. I've been sitting with that response for a few days. Eighty percent of CEOs recently admitted they've seen no productivity gains from their AI investment. Hundreds of billions spent globally. Boards briefed. Strategies signed off. Budgets approved. By whom? Now those same boards are preparing to spend even more. The fear of being left behind is real, and the vendors selling the solution are ready for it. But before the next wave of investment gets approved, one question is worth asking. Are you genuinely transforming how your people work - or are you applying AI lipstick to a tech stack that's been failing you for decades? Legacy technology doesn't accumulate by itself. But in large organisations, every new CTO inherits the debt and blames the last one. The buck has been passing for years. While they're pointing fingers, a new generation of companies isn't waiting for the post-mortem. AI-native businesses are being built right now - faster, leaner, with much lower operating costs. Better products. And people who are redefining what's possible instead of explaining why it isn't. In the next two years, 80% of CEOs might just find out that you don't get to own the investment decision and disown the outcome.
14 Comments
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top contentOthers named Mathew Dostal
2 others named Mathew Dostal are on LinkedIn
See others named Mathew Dostal