Notes on technical judgment.
The mistakes early companies make are remarkably predictable. I write down the ones I keep watching founders walk into — and how to see them coming. No theory, just patterns from the room.
Your AI wrote the code. Nobody wrote down why.
When an engineer writes your code, the reasoning lives in their head and you can ask. When an agent writes it, the reasoning often lives nowhere.
DecisionsYour AI vendor can change its price and move your margin
Your product runs on an AI vendor's API, so your gross margin is a price you do not set. Here is how to price the repricing risk before it hits.
HiringCan you manage an engineer more technical than you?
You cannot check the code and cannot judge an estimate. Here is how to manage a first engineer who is far more technical than you are.
HiringYour engineer search hit three months. Lower the bar?
The search is in month three and the seat is still empty. Before you lower the bar on your first engineer, find out which problem you have.
HiringHire your first engineer on a two-week paid trial
You will not know if an engineer is good until you watch them work for two weeks. That is what a paid trial buys you for your first hire.
HiringYour first engineer wants to work remote. Say yes?
The best candidate for your first engineering role will not relocate. Here is how to decide whether to let hire number one work remote.
HiringShould you use a recruiter for your first engineer?
A recruiter wants 22 percent of salary to fill your first engineering seat. Here is when that fee is worth paying, and when it is not.
Pattern recognitionYour team pastes your code into ChatGPT. Who wrote the rule?
Your team already pastes code, customer data, and contracts into personal AI accounts with no rule for it. Banning the tools fails. Here is the one-page rule that holds.
DecisionsFlying blind in prod? When to pay for observability
You only notice monitoring when something breaks. The three cheap tools that stop the guessing, and when a full platform is worth its bill.
DecisionsDo you need feature flags yet, or just a deploy?
Feature flags are high leverage for small teams, but the version you need at five engineers is not the enterprise plan being sold to you.
DecisionsYour app is slow. Add a cache, or fix the query?
A cache hides a slow system, it does not fix one. How to tell a slow query from an over-called one, and when a cache is actually the right tool.
DecisionsYou built the MVP on no-code. When do you leave?
No-code got you launched. The concrete signals that you have hit the ceiling, and how to run the migration as a money decision, not a vibe.
DecisionsDo you need Kubernetes yet? Probably not
Most early teams adopt Kubernetes for a scaling problem they do not have. The signals that mean you actually need it, and what to run first.
Knowing whenWhen to freeze engineering hiring on purpose
A hiring freeze is a decision, not a failure. Here is how to know when to pause engineering hires, and how to do it without wrecking morale.
Knowing whenWhen to stop outsourcing and build in-house
Outsourcing your build is right until it isn't. The switch is not about cost. It is about control, knowledge, and whether the product is now your core.
Knowing whenWhen one engineer isn't enough anymore
The jump from one engineer to two is not about workload. It is about bus factor, review, and whether your whole product lives in one person's head.
Knowing whenWhen to end your fractional CTO engagement
Most fractional engagements should end on purpose, not drift. Here are the signals that the work is done and you are paying for a habit.
Pattern recognitionYour AI coding agent is choosing your supply chain
An AI coding agent does not just write your code. It picks your dependencies, and it skips the check a careful engineer would run. Here is the risk, and the fix.
HiringYou can't tell if you're interviewing the engineer or their AI
Any candidate can now have an AI feeding them answers mid-interview, and you will not spot it. Stop trying to catch the cheat. Run an interview AI cannot rescue.
DecisionsYour AI bill is climbing. Switch to an open-weight model?
The price gap between frontier APIs and open-weight models is real. But switching moves cost off an invoice you can read and onto engineering time you cannot.
FundraisingYour incident history is a diligence document now
Reviewers no longer just ask if you have outages. They ask what happened last time and what you changed. How you handled the last fire is the signal.
VendorsOne vendor can take your whole product down
Somewhere in your stack is a vendor you cannot live without and have no plan to replace. Diligence looks for it, and so should you, before it fails.
FundraisingYour database dies at 2am. Can you get it back?
Backups you have never restored are not backups. Diligence asks how fast you recover and how much you lose, and a vague answer is the finding.
FundraisingThe term sheet is signed. Diligence isn't over.
The term sheet is not the finish line. Confirmatory diligence comes after, and it is where a solid-looking deal can still quietly fall apart. Here is what to expect.
FundraisingYour roadmap is a wishlist. Diligence will notice.
Investors do not just read your roadmap for the features. They read it for evidence you can decide what not to build. Here is what a credible one looks like.
Pattern recognitionEvery tool you connect to your AI agent taxes every call
Every tool you connect to your AI agent loads into the model's context on every request, used or not. That silent tax hits cost, latency, and quality at once.
DecisionsYour AI agent is live and failing. Pull it or fix it in place?
A live AI agent that just did the wrong thing in front of a customer is not a debugging problem first. It is a containment decision. Here is the order to make it in.
VendorsYour agency subcontracted your product. You would not know.
An agency can only assign IP it actually owns. If a shop you have never heard of wrote your payments module, the chain of title has a link nobody has checked.
VendorsFixed bid, dedicated team, or staff aug: picking wrong
Founders shop agencies on price and portfolio, then accept whatever engagement model is put in front of them. That model decides how the next nine months go.
VendorsYour agency says it is done. Done by whose definition?
Vague scope drives most outsourcing disputes. Not because agencies are dishonest, but because nobody wrote down how anyone would know the work was finished.
VendorsYour agency got faster with AI. Your bill did not move.
Time and materials assumes an hour is a decent proxy for an outcome. That assumption is under strain, and your contract never got the memo.
VendorsYour agency AI-wrote your code. Who owns it?
Your MSA assigns work the agency creates. Nobody defined what creates means when a model writes the first draft and a contractor approves it in ninety seconds.
VendorsThe senior team that pitched you isn't building your product
The senior engineer whose name is on the proposal gets replaced, four to eight weeks in, by a mid-level developer with a senior title. You are still paying the senior rate.
VendorsYour agency is done. Can your team take over the code?
The contract ends, the agency rolls off to the next client, and what you are left holding is a black box. Nobody planned the handoff that would have prevented it.
VendorsYour agency sends busy status reports. Is anything shipping?
Status reports full of activity are the easiest thing for an agency to produce, and the easiest thing for a non-technical founder to mistake for progress. Motion is not movement.
VendorsTwo agency bids, one triple the other. Which is right?
One agency quotes 40,000 dollars, another 130,000 for the same brief. The spread is not a mistake. It is two shops reading your one-paragraph scope in different ways.
VendorsThe SOW clauses that cost you six months later
A statement of work decides, in advance, who eats the cost when a project goes sideways. The expensive clauses are never hidden. They sit in the sections you skip.
FundraisingAn AI wrote the diligence questions your investor will ask
A growing share of investors run first-pass diligence through an AI agent that reads your data room and asks sharper, more thorough questions than a junior associate ever did.
DecisionsHow much testing do you need before you have users?
The advice says 80 percent coverage from day one. Before product-market fit that is mostly wasted effort. Where the little testing you do write should actually go.
DecisionsYour engineer wants to adopt a new tool. Say yes?
Your engineer wants a new database, framework, or Kubernetes. Five plain-language questions to tell a real need from resume-driven development, no code required.
DecisionsA customer wants an API. Should you build one?
One customer asks for an API and it sounds like a small yes. A public API is a product you maintain forever. How to decide, and the lighter options to try first.
DecisionsA customer wants to self-host. Should you let them?
A big customer says the product must run in their own cloud. Self-hosted, single-tenant, and BYOC are not the same, and the wrong yes can reshape your roadmap for years.
DecisionsDo you actually need a native mobile app?
Most founders put a native mobile app on the roadmap too early. A good web experience covers most of the need at a fraction of the cost. When native is actually worth it.
Pattern recognitionYour AI agent can run all night. What stops it?
A shipped feature that breaks fails once. An AI agent that breaks keeps acting until something stops it. Most teams built no circuit breaker.
Pattern recognitionYour AI agent keeps guessing what you meant
Your AI agent builds the wrong thing because you left the decisions unspecified and it guessed. A short spec, written before it starts, is the fix.
Knowing whenWhen to hire a full-time CTO before product-market fit
The standard advice is to wait for product-market fit before a full-time CTO. For a specific kind of company that advice is wrong and waiting is the expensive mistake.
Knowing whenA technical advisor might be all you need, not a CTO
Between doing it all yourself and hiring a fractional CTO sits a lighter option founders skip: a paid technical advisor. Here is when that is the right call.
Knowing whenWhen to hire technical leadership before your first engineer
Most founders hire builders first and a leader later. Sometimes that order is backwards and it costs you a rewrite. Here is how to tell which case you are in.
Knowing whenYour fractional CTO won't tell you to hire full-time
The person best placed to tell you it's time for a full-time CTO is the one whose retainer ends when you do it. Here is how to read the signal yourself.
Pricing the workWhen is a specialist fractional CTO worth the premium?
Fintech, healthtech, and AI fractional CTOs charge 20 to 40 percent more. Sometimes that premium saves you far more than it costs, and sometimes a generalist is the smarter buy.
Pricing the workYour fractional CTO asked for a rate increase. Is it fair?
A year in, your fractional CTO wants to raise their rate. How to tell a fair adjustment from a squeeze, and what to check before you say yes or push back.
Pricing the workPay your fractional CTO upfront or in arrears?
Monthly upfront, in arrears, deposits, net-30: the payment terms on a fractional CTO deal matter more than founders expect. Here is what protects you and what protects them.
Pricing the workYour fractional CTO blew past the retainer. Now what?
The month-four invoice is bigger than the retainer, and nobody warned you. Here is why overage happens, what a fair overage looks like, and how to fix the setup.
Pricing the workThe day rate is a bad way to compare fractional CTOs
Two fractional CTOs quote the same day rate. One can quietly cost you three times as much. Here is what to compare instead of the number on the first line.
Pattern recognitionYour AI agent plugs into tools nobody vetted
Your AI agent connects to external tool servers you did not build or vet. Each is a dependency and a channel to steer your agent. Here is how to vet them.
FundraisingYour privacy policy promises what your code does not do
Your privacy policy says one thing about customer data. Your actual data flows do another. In diligence and in a breach, that gap is the liability, not the code.
FundraisingHow deep technical diligence goes at seed vs Series A
At pre-seed almost nobody looks at your code. By Series A a stranger reads your architecture for a week. Knowing which stage you are in tells you exactly how much to prepare.
FundraisingWhen your engineer contradicts you in the diligence call
You told investors the architecture scales. On the diligence call, your lead engineer says the opposite. That gap does more damage than the flaw itself.
FundraisingThe secrets in your git history diligence will find
You deleted the API key from your code. It is still in your git history, and a technical reviewer runs one command that surfaces it. Here is how to clean up before they look.
FundraisingWhere did your training data come from? Diligence asks now
The first question a technical reviewer asks an AI startup in 2026 is not about your model. It is where your training and eval data came from, and whether you can prove it.
FundraisingYour gross margin is a technical question in diligence
Investors read your gross margin as a verdict on your architecture. Below 75%, they dig into your infrastructure bill. Here is what they find.
FundraisingThe deck number diligence will make you prove
Every technical claim in your pitch - 99.9% uptime, sub-second latency, real-time - becomes a diligence request for evidence. Here is how to be ready.
FundraisingCan it handle 10x? What that diligence question means
Investors ask if your system handles 10x growth. They are not asking you to have built it. Here is what the question is really testing.
FundraisingDiligence found a real problem. What happens next?
A real diligence finding rarely kills the round. It reshapes the deal. Here is what actually happens and how to steer it.
Pattern recognitionYou shipped an AI feature. How do you know it is good?
Your AI feature demoed well and shipped. But how do you know it stays good after a prompt change or model swap? Evals are the cheap answer.
HiringAI agents do the junior work. Should you still hire juniors?
Agents now do what junior engineers used to. That does not mean you stop hiring them. It means you hire them for review bandwidth and pipeline, not boilerplate.
HiringThe questions a great engineer asks you back
When you cannot read code, the strongest interview signal is not their answers. It is the questions a senior engineer asks you, unprompted, at the end.
HiringYour first engineer will get bored around month 18
The moment your first engineer becomes indispensable is the moment you are most at risk of losing them. The usual exit is boredom, not money.
HiringYour first engineer job post is doing the wrong job
Most first-engineer job posts read like a big-company role and pull the wrong people. The post is a filter for a kind of person, and a sales document.
HiringThe senior engineer said yes. Then the counteroffer landed.
In 2026 about 60% of senior engineering offers get countered. Here is how to manage the offer stage so you do not lose the hire in the last mile.
HiringYour first engineer starts Monday. Now what?
Founders treat the offer as the finish line. Then the first engineer sits down, and a badly run first 90 days quietly wastes the person you fought to land.
Pattern recognitionYou gave your AI agents the keys. Nobody scoped them.
The AI agents in your stack are holding broad, long-lived credentials nobody scoped. Here is how that goes wrong at pre-seed to Series A, and the fix.
DecisionsYou deploy straight to prod. Do you need staging yet?
Pushing to production is fine at two users. Here is the line where it stops being lean and starts being an uninsured bet.
DecisionsThe project you should abandon, not finish
Every team has a project everyone privately knows is dead, that nobody kills because too much has gone into it. How to walk away.
DecisionsUpgrade the dependency now, or pay more for it later?
Deferring a version upgrade is free this week. The bill for years of deferred laters arrives at once, usually during a crisis.
DecisionsFreeze features and fix reliability, or keep shipping?
Your velocity looks great and your error rate is climbing. How to know when to stop adding features and start hardening.
DecisionsSelf-host it or pay for managed? Run the real cost math
The managed database looks 10x more expensive than a $30 server. Then you count the engineering hours self-hosting quietly eats.
DecisionsThe AI model your product runs on will be retired
The model behind your AI feature is on a clock. Vendors retire models every few quarters, and when yours sunsets you migrate on their schedule. Here is how to carry that risk.
Pattern recognitionThe shared database that couples your whole team
One database is fine early. No boundaries inside it is the smell that makes every later change slow and risky. How to spot it and fix it cheaply.
Pattern recognitionYou keep polishing instead of launching
The team that never launches looks responsible and is quietly avoiding the one verdict that matters. How to tell polish from fear, and ship.
Pattern recognitionThe prototype that quietly became your production system
The most load-bearing code in many startups was a weekend prototype nobody meant to ship. How that promotion happens and what it costs you later.
Pattern recognitionYou hired a team before you had a product
Hiring a whole team before the product earns it is the pattern behind most early-stage layoffs. How to spot growth theater and what to do instead.
Pattern recognitionThe second product you built before the first one worked
Starting a second product before the first one retains users is one of the most expensive founder mistakes I watch happen. Here is the pattern and the cost.
Pattern recognitionThe security holes hiding in your AI-written code
AI coding tools ship insecure code by default. If nobody with security judgment has read yours, assume there are holes, and here is how to find them.
Knowing whenWhen fractional leadership is the wrong call from day one
Fractional leadership fits most early startups, but not all. The cases where you need a full-time technical leader from day one.
HiringShould your founding engineer get the CTO title?
Your first engineer wants the CTO title. Why the title and the job differ, and how to decide without a painful reversal later.
Knowing whenThe events that mean you need engineering leadership now
Headcount is a poor trigger for hiring engineering leadership. These are the events that tell you the cost of waiting has arrived.
Knowing whenYour investors want you to hire a CTO. Should you?
Investor pressure to hire a full-time CTO is often pattern-matching, not a real signal. How to tell the difference and what to do.
Knowing whenWhen to stop looking for a technical co-founder and hire
The co-founder search becomes avoidance at some point. Here is when to stop looking and start building with hired talent instead.
Pricing the workWhat a one-time technical review should cost
You do not need a full-time CTO to answer one urgent question about your code or your team. You need a review. Here is what one costs and when it is worth buying.
Pricing the workMilestone billing or a monthly retainer for tech work?
Pay per deliverable, or pay a flat fee every month? The right structure depends on whether you are buying output or buying judgment. Here is how to tell.
Pricing the workFractional CTO or dev agency: which is cheaper?
On paper the hourly rates look almost identical. The real cost difference is buried in markups and in what you are actually buying. Here is the honest math.
Pricing the workHow to compare two fractional CTO proposals
Two fractional CTOs, two proposals, wildly different prices, and you cannot read code to tell them apart. Here is what to compare instead of the number at the bottom.
Pricing the workWhat a fractional CTO retainer actually buys each month
You send a flat monthly fee and get back... what, exactly? Here is the breakdown of what is inside a fractional CTO retainer, what is out, and how to check you are getting it.
HiringDo you need a forward deployed engineer, or a founding one?
Everyone says hire a forward deployed engineer. Here is what the role really solves, why most seed founders reach for it too early, and what to hire instead.
FundraisingWhat belongs on the technical slide of your deck
Founders either skip technology in the deck or drown one slide in an architecture diagram nobody reads. The technical slide has a job, and it is not to prove you can draw boxes.
FundraisingThe engineering hiring plan investors actually read
Your use-of-funds slide says sixty percent engineering and stops there. Investors read the hiring plan behind that number more closely than you think. What makes it credible.
FundraisingHow long technical diligence takes, and why it stalls
A term sheet is not the finish line. Technical diligence runs four to eight weeks, and the delays are almost never the analysis. They are the documents you did not have ready.
FundraisingThe open-source license that can sink your valuation
Modern apps pull in hundreds of dependencies, and one copyleft license buried three levels deep can become a diligence problem that costs you leverage. Find yours first.
FundraisingDo you need SOC 2 to raise? Usually not yet.
An investor asks if you are SOC 2 compliant and you hear a requirement. It usually is not one. Here is how to tell the signal from the badge, and what to do if you are not there.
DecisionsThe AI coding tool bill nobody budgeted for
Your team's AI coding tools run from a few hundred to a couple thousand dollars per engineer a month. Here is how a founder tells worthwhile spend from waste.
FundraisingThe VC sent a technical advisor to diligence you
The investor liked the deal, then sent a technical advisor to check your codebase. How that session works and how to walk in ready.
FundraisingShould you tell investors about your technical debt?
Every company has technical debt and investors know it. The question is not whether to disclose, but how to make it read as judgment.
FundraisingDo you have an AI moat, or are you just renting one?
When AI makes products cheap to build, investors fund the moat, not the product. How to tell whether you own one or just rent the model.
FundraisingThe IP paperwork that quietly stalls your term sheet
Missing IP assignments are the number one reason a signed term sheet freezes. How to build a clean ownership trail before you raise.
FundraisingThe technical half of your data room is probably empty
The finance half of your data room is polished. The technical half is a folder someone started and never finished. Here is what belongs in it.
Pattern recognitionYour AI agent demo wowed the room. Can it ship?
A demo has to work once, with you watching. Production has to work on the ten-thousandth input, with nobody watching. That gap is where most AI agent startups quietly stall.
Pattern recognitionYour AI team feels fast. Is it actually shipping faster?
AI makes a team look busy: more commits, more pull requests, more demos. That is not the same as shipping faster. Here is how a founder tells real velocity from motion.
HiringWho vets your engineer when you have no engineers?
You cannot read code and your first hire has no technical peers to interview them. Renting technical judgment for a few hundred dollars an interview beats guessing and hoping.
HiringWhere to actually find your first senior engineer
Job boards return volume, not senior engineers who want a zero-to-one role. The people you actually want are already building somewhere. Here is where they are.
HiringThe take-home test stopped working. Do this instead.
AI rebuilt the floor of what a take-home proves, and 71% of engineering leaders now say technical skills are harder to assess. A short paid work sample gives you the signal back.
HiringHire a generalist first, not a specialist
A specialist hired too early sits idle. A generalist hired for a specialist role drowns. For your first engineer, breadth beats depth until you have a real bottleneck.
DecisionsThe security questionnaire that stalls your biggest deal
A 200-question security review lands mid-deal and your biggest contract goes quiet. The fix is separating the controls you can prove today from the badge that takes months.
DecisionsWhen to force your team onto one stack
Five engineers, four languages, three ways to deploy. At some point the sprawl costs more than the freedom. Standardize too early and you optimize a problem you do not have yet.
DecisionsShould you sign a cloud commit to cut the bill?
Cloud providers will happily cut your bill by up to 72% if you commit for a year. That discount is a bet on a forecast you probably cannot make yet. Here is how to size it.
DecisionsThe feature you should kill, not maintain
Not every broken feature deserves a fix. Some deserve a funeral. Deleting the right feature is one of the highest-leverage technical decisions a small team can make.
DecisionsShould you build your own login, or just buy it?
Almost every early team that rolls its own authentication regrets it by year two. Here is the short list of reasons to build, and the much longer list of reasons to buy.
VendorsYour fractional CTO now comes with an AI agent team. Buy it?
More fractional CTO pitches now bundle an AI agent team that ships code. Sometimes that is real leverage, sometimes repackaged agency output. Here is how to tell them apart.
Pattern recognitionThe integration sprawl nobody decided to build
No one chose to run fifteen SaaS tools stitched together with webhooks and glue code. It accrued one reasonable decision at a time, and is now the most fragile part of the company.
Pattern recognitionThe early tech decisions you can't cheaply undo
Most early technical choices are easy to reverse, so move fast. A handful are one-way doors that get exponentially expensive to change. Knowing which is which is most of the job.
Pattern recognitionThe refactor you keep promising never arrives
Every team says they will clean it up after launch. Most never do. The difference between debt you pay down and debt you default on is a pattern worth recognizing early.
Pattern recognitionYour biggest customer is writing your roadmap
When one large account's requests keep topping the backlog, you are not building a product anymore. You are building a custom system for one client and calling it a roadmap.
Pattern recognitionAdding engineers to a late project makes it later
When a deadline slips, the instinct is to add people. In the short term it almost always pushes the date further out. Here is why, and what to do instead.
Pattern recognitionYour AI agents write faster than anyone can review
AI agents generate code faster than anyone can read and vouch for it. The bottleneck moved from writing code to reviewing it, and most teams optimize the wrong part.
Knowing whenYour team ships fast and builds the wrong things
The engineers are productive. Tickets close. And somehow the product keeps drifting. That gap between output and outcome is usually a leadership signal, not a hiring one.
Knowing whenDo you need a fractional CTO for 3 months or forever?
Some technical needs are a one-time decision. Others need someone to stay and own the outcome. Confusing the two is how founders overpay or underbuy.
Knowing whenWhen every technical decision waits on you
You are not technical, but every technical call still routes through you. That is a signal, not a workflow. Here is how to read it and what it usually means.
Knowing whenIs it too early to hire a fractional CTO?
Most writing about fractional CTOs tells you when to hire one. Here is the harder question almost nobody answers: when even a fractional one is too early.
FundraisingAI wrote your code. Can it survive diligence?
AI-generated code can pass the demo and fail the diligence read. Reviewers grade who understands the system and what it costs to own, not how fast you shipped.
Pricing the workFractional or full-time CTO? Run the numbers first
A full-time CTO costs $400k to $500k all-in plus equity; a fractional one runs about $120k. Run the real math before you decide which you need.
Pricing the workHow to budget for technical leadership before a CTO
Before you can afford a CTO, technical leadership is a ladder of four rungs from advisory hours to full-time. Here is what each rung costs and buys.
Pricing the workThe hidden cost of a cheap fractional CTO
A $2,500 a month fractional CTO looks like a bargain until a bad architecture call or wrong hire arrives. Here is what the discount really costs.
Pricing the workWhat a fractional CTO actually costs in 2026
Fractional CTO quotes run from $2,500 to $30,000 a month for the same title. Here are the honest 2026 ranges and how to tell which job you are buying.
FundraisingNo technical co-founder. Will it sink your raise?
Investors do not require a technical co-founder. They require an answer for who owns the technology and whether it can be built.
FundraisingAn agency built your product. Diligence will find out.
Outsourcing the build is fine. Being unable to prove you own it, or explain it, is what turns diligence into a problem.
FundraisingThe technical diligence red flags that kill a round
Most rounds do not die over messy code. They die over a handful of specific, findable problems. Here are the ones to fix first.
FundraisingHow to explain technical risk to your board
Boards do not punish you for technical risk. They punish you for surprises. Here is how to brief them so risk reads as control.
FundraisingThe 3-page tech memo that gets you through diligence
Write a short technical memo before anyone asks for one. It changes the tone of diligence from interrogation to a guided tour.
Pattern recognitionYour AI-built app will hit a wall. The question is when.
Apps shipped fast with AI tend to hit a wall a few months in, when the codebase stops absorbing change. The wall is predictable, and so are the signals.
HiringThe reference check questions that actually tell you something
Most engineering reference checks are theater: a friend confirms the candidate is great. Here are the questions that get past the script and the silences worth reading.
HiringWhat to pay your first engineer: salary and equity
Founders agonize over the salary number and wave at equity, which is backwards. Here is how to think about both for your first engineer, with real 2026 ranges.
HiringHow to interview a senior engineer when you can't read code
You can't grade code you can't read, but you can grade judgment. Here is how a non-technical founder runs a senior engineering interview that actually filters.
HiringWhen does your dev team need its first lead?
Founders reach for an engineering lead by headcount, usually too early or too late. The better trigger is pain: standups that drag and work that stalls on dependencies.
DecisionsYou have outgrown a tool. Is it worth the cost of leaving?
Switching a core tool or platform always costs more than the demo suggests. Here is the threshold test I use to decide migrate now, later, or never.
DecisionsMonolith or microservices? The decision is about your team, not your code
Founders ask whether to go microservices for scale. Almost always the answer is a monolith, and the real trigger to split is org size, not traffic.
DecisionsEverything feels broken. How to triage technical debt that matters
When the whole codebase feels like debt, founders freeze or overspend. Here is the triage I use to decide what to fix now, later, or never.
DecisionsRewrite or refactor? How to make the call without betting the company
Most rewrites overrun by 2-4x and many never ship. Here is the test I use to decide whether a codebase needs a rewrite or just disciplined refactoring.
Pattern recognitionYour AI feature might be losing money on every user
An AI feature can lose money on every user without anyone noticing, because its cost is usage-based and shows up late. Find your cost per action first.
Pattern recognitionWhen your whole codebase lives in one person's head
One engineer who understands everything feels like a strength until the day they are unavailable. Key-person risk is the early smell founders ignore most.
Pattern recognitionPremature scaling: the pattern that quietly kills startups
Premature scaling is the most common way early startups die, and it rarely looks like scaling. Here is what the pattern looks like up close.
Pattern recognitionExpensive mistake or expensive necessity? How to tell
Every founder eventually faces a six-figure technical decision. Here is the pattern I use to tell a costly necessity from a costly mistake.
Pattern recognitionThe architecture smells I look for in a 20-minute review
A handful of signals predict whether your next six months will be fast or slow. Here are the ones I check before reading a single feature.
DecisionsBuilding an AI agent? You may be building the wrong half
Founders frame an AI feature as build or buy. The real question is which thin layer is actually your moat. Build that, rent the rest.
Knowing whenWhen the technical founder should stop being the CTO
Technical founders blow past the moment they have become the engineering bottleneck by a year or more. Here are the signals, and what to hand off first.
Knowing whenMoving from a fractional to a full-time CTO cleanly
You decided you need a full-time CTO. The handoff from your fractional leader is where a year of hard-won context quietly leaks away, unless you plan it.
Knowing whenYou need a developer, not a CTO yet
Many founders say 'I need a CTO' when they need someone to build the first version. Confusing the two costs you a quarter and a chunk of your cap table.
HiringCTO or VP of engineering: which do you hire first?
Founders ask this as a titling problem. It is not. The two roles fix different failures, and hiring the wrong one is expensive in a way you will not feel for months.
Knowing whenSigns you've outgrown your fractional CTO
A fractional CTO is right for a season, not forever. Here are the concrete signals that you have outgrown the half-day cadence and need someone full-time.
HiringIf AI writes the code, what do you hire an engineer for?
AI made writing code cheap. It did not make judgment, review, or ownership cheap. Here is what your first senior engineer is actually for now, and how to interview for it.
FundraisingYou paid for the code. You might not own it.
You paid the invoice, but the rights may still belong to the agency, contractor, or co-founder who wrote the code. That gap is what stalls a round.
Pricing the workShould you pay a fractional CTO in equity?
Cash-tight founders often offer a fractional CTO equity instead of fees. Sometimes it is smart. Often it is a quiet way to misalign both sides.
Pricing the workHow many days a month does a fractional CTO need?
Founders ask how many days they should buy. Days are the wrong unit. Here is what actually drives the number, and the ranges I see in practice.
Pricing the workWhat senior engineering judgment is actually worth
AI writes the code now, so why does a senior still cost three times a junior? Because you are paying for the decisions, not the typing.
Pricing the workScope your fractional CTO before it becomes full-time
The biggest cost in a fractional engagement is not the rate. It is the quiet creep from two days a week to five, billed one urgent ask at a time.
Pricing the workHow to tell if your fractional CTO is worth the money
A fractional CTO costs $8k to $15k a month. Here is how to tell, inside 90 days, whether you are buying outcomes or just buying activity.
DecisionsYour AI-built MVP is a draft, not a product
Built your MVP with AI and it works? Do not rewrite it. The risk is not the code quality, it is that nobody can explain or safely change it.
VendorsDon't sign the big contract. Run a pilot first.
Most founders sign a year-long agency contract on a sales call and a portfolio. A short paid pilot tells you what neither can.
VendorsHow to cut your agency bill without firing them
Switching agencies is costlier than the new quote shows. Here's how to cut the bill without losing the team that built your product.
VendorsManaging an offshore dev team when you're not technical
An offshore team amplifies whatever management you bring to it. Here's how to run one well when you can't read the code yourself.
VendorsThe fixed bid that quietly becomes 60% more
A fixed-price contract only protects you if the scope is fixed too. Here's how a $120k quote quietly becomes $190k, and how to stop it.
DecisionsGood technical debt: when the shortcut is the right call
Not all technical debt is a mistake. The best early teams take it on deliberately, write down the terms, and repay it before it compounds.
Knowing whenDo AI coding agents change when you hire your first engineer?
AI coding agents ship real work in 2026. That moves the timing of your first engineering hire more than it removes the need for one.
HiringYour first engineer doesn't have to be an employee
Founders open a full-time role by default. Often a contractor, a fractional lead, or a small bench buys more runway and keeps you reversible.
NotesWelcome to the new writing setup
A short note on how this site is now published — and why I'd rather press 'Publish' than commit HTML.
HiringYour first senior engineer is a bet. Here's how to stop guessing.
The wrong first technical hire sets you back a year. The screen that actually predicts whether someone can lead — not just code.
Pricing the workWhy I don't bill by the hour
The second you're watching a clock, you stop calling me with the decisions that actually matter. The case against the timesheet.
FundraisingWhat investors actually ask about your architecture
The technical due-diligence questions that stall rounds — and the three-page document that answers all of them.
Knowing whenYou're not ready for a full-time CTO. Here's how to know.
Three signals you've outgrown fractional — and an honest test for whether technology has become the whole company.
VendorsThe agency invoice nobody reads
How to spot when you're paying for motion instead of progress — line by line, in the bill you've been rubber-stamping.
10 technical mistakes that kill seed-stage startups.
A one-page field guide, free. The predictable, expensive, irreversible ones — and the early signal for each. Plus the occasional note when I write something new.