// Blog
Notes on building for the long clock.
Essays on building AI-native SaaS inside a studio, unit economics, and what we're learning building our brands.
39 of 39 essays
Patient capital, not permanent capital: how we think about exits now.
Cobalt Glacier is patient by default, outcome-driven per brand — indefinite hold, strategic sale, recap, or eventual public listing. Why we shifted from permanent-capital language to patient-capital discipline, and what it means for founders and investors.
Security posture on a long hold: three programs, not a compliance event.
Identity and access, data lifecycle, and incident response are three concurrent programs that produce compliance as a byproduct. The framework Cobalt Glacier uses to retune security posture after close — and why it ends up being a topline lever.
Capital allocation inside an AI venture studio.
Reinvest, build a new brand, reduce debt, distribute. The four legitimate uses of portfolio cash flow, the default ordering Cobalt Glacier holds to, and the annual written argument that keeps allocation decisions honest over decades.
Tuck-in versus new platform brand: how we decide which a deal is.
A real tuck-in requires customer, workflow, and data model overlap with the parent brand. Anything less is a new brand pretending to be a tuck-in. Cobalt Glacier's default is new platform — we'll protect operator continuity at the cost of headline synergies.
Founder stamina on a twenty-five-year hold: removing the exit clock.
Permanent capital removes the exit clock most founders have been operating against. The structured renegotiation of role, calendar, team, and compensation in the first year is what makes the long arc sustainable.
Open source dependency risk in B2B SaaS diligence.
License risk is the smallest of three. License drift, maintainer collapse, and supply-chain compromise are the underwriting questions that matter on a twenty-five-year hold. The Cobalt Glacier dependency workstream and remediation pattern.
Per-seat, usage, or platform fee: choosing the right SaaS pricing meter.
The pricing meter encodes a theory of value. Most growing B2B SaaS brands are charging on a meter that no longer maps. The Cobalt Glacier diagnostic and the remap pattern we use in the first four quarters after launch.
Board composition for a permanent-capital portfolio brand.
Three seats, an operator chair, a fixed quarterly cadence, and one job — protecting the long-hold thesis against short-term operating pressure. The brand-level board is the most important governance forum in the portfolio.
Channel versus direct in vertical B2B SaaS: a four-question filter.
Channel partnerships compound when the partner owns trust the brand cannot replicate. They leak economics when they don't. The Cobalt Glacier filter: who owns the trust, the implementation, the renewal, and the data.
Customer support is a margin and retention lever, not a cost center.
Run as a cost center, support erodes both margin and NRR. The Cobalt Glacier three-job framework — friction reduction, product feedback, renewal protection — typically moves support from 12–15% of revenue into the 7–9% band within four quarters.
Why we became a studio.
Cobalt Glacier started out writing an acquisition thesis. It never closed a deal under it. Here's the honest account of why we build our own AI-native brands instead.
The three SaaS quality-of-earnings adjustments standard QoE misses.
Engineering investment, deferred revenue mechanics, and CAC economics are the three QoE adjustments most often missed in lower-middle-market B2B SaaS — and the three that most change the steady-state earnings number a long-hold buyer will own.
Engineering productivity on a twenty-five-year clock.
Velocity is the wrong instrument. On a permanent-capital hold, engineering productivity is measured by change-failure rate, lead time to a paying customer, share of compounding work, and incident minutes per active customer.
International expansion on a twenty-five-year clock.
Most B2B SaaS international expansion fails because it is underwritten as a sales expansion when it is a market-development commitment. The Cobalt Glacier playbook: pull signal first, country lead before AE, retention before bookings.
Resolving a co-founder buyout before a brand takes on a long-term outcome.
Unresolved co-founder cap-table situations are the most common reason a clean B2B SaaS process stalls. The fix is a documented buyout twelve to eighteen months before the studio conversation, not during it.
Proprietary data is the only real AI moat in vertical B2B SaaS.
Model access is not a moat. Cobalt Glacier builds AI features on the data behind them — workflow-generated, non-obvious schema, closed feedback loop — not the model in front of them.
Integration anti-patterns: what we deliberately do not consolidate across the studio.
Brand consolidation, product surface merging, sales pooling, support pooling, and unified billing — the five integration anti-patterns Cobalt Glacier refuses, and the willingness test that keeps the platform-brand line honest.
Gross margin floors in AI-native B2B SaaS: how we underwrite the cost stack.
Cobalt Glacier's 70% steady-state gross margin floor for AI-native brands, the three axes that determine where the floor sits, and the patterns that disqualify a brand idea regardless of topline growth.
What being CEO of a Cobalt Glacier brand actually looks like.
Cobalt Glacier's default assumption is that the founder stays as CEO indefinitely. What changes after close — a real board, a long-horizon capital allocator, a shared services platform, and a peer group of operators — and what deliberately does not.
How we migrate pricing after a new brand launches without breaking NRR.
A four-quarter pricing playbook — instrument, repackage, fix contracts, then move list — designed to capture unrealized economics in a growing SaaS brand without spending the trust the brand was built on.
Customer concentration is the most underpriced risk in lower-middle-market SaaS.
How Cobalt Glacier underwrites customer concentration on three axes — revenue share, contractual entrenchment, and narrative dependence — and the diligence workstream we actually run.
Building a shared services platform for an AI venture studio.
The seven functions on the Cobalt Glacier platform, the four we deliberately keep off it, and the single test — would the brand willingly buy it from us at market price — that decides whether a service belongs on the studio platform.
Why we pass on most venture-backed B2B SaaS operators as co-founders.
We pass on most venture-backed SaaS operators not because they are bad operators, but because a venture capital stack is mismatched with a patient-capital building period. The structural reasons, and the narrow conditions under which we lean in.
The Cobalt Glacier B2B SaaS brand-readiness checklist.
Six readiness workstreams — commercial, product, customer, financial, legal, and operational continuity — and the exact items an operator co-founding a brand with us can prepare to compress the timeline without compressing the rigor.
Pricing power in vertical SaaS: why narrow workflows compound faster.
Pricing power is the right to raise prices annually without measurable churn. Vertical B2B SaaS earns it faster than horizontal SaaS because the workflow is narrower, the competitor set is shallower, and the buyer feels the pain directly.
Multi-product B2B SaaS: when bundling compounds, when splitting is the honest answer.
Bundling B2B SaaS products only compounds when the buyer, the workflow, and the data model overlap. The Cobalt Glacier framework for deciding when to bundle a roadmap, when to split it, and when to bundle only the buying motion.
Free cash flow conversion is the underwriting bar in B2B SaaS.
Why we underwrite every Cobalt Glacier brand to a steady-state seventy-five percent free cash flow conversion ratio, the four reconciling items we focus on, and what disqualifies a brand idea from our process.
How the studio actually pays co-founding operators: equity, milestones, and ongoing economics.
Cash sized to remove financial pressure, meaningful equity in the brand, an operator-controlled milestone structure, and ongoing operator compensation — the four building blocks of a Cobalt Glacier co-founding arrangement and the kind of operator it fits.
LP-grade reporting for a patient-capital AI venture studio.
What we publish quarterly and annually, the governance stack behind the reporting, and why we hold ourselves to an institutional reporting standard nobody is making us follow.
Concentration is the strategy: why a five-brand portfolio is the right shape.
Why Cobalt Glacier holds five B2B SaaS brands on purpose, what concentration buys us that diversification cannot, and the specific conditions under which we will add another.
How we underwrite a new brand for a 25-year hold.
Three filters every brand idea has to clear before we greenlight it, what changes when the hold period is twenty-five years, and why the discipline is in the ideas we walk away from.
The first 100 days: what we set up when a new brand joins the studio (and what we deliberately don't).
The playbook we run when a new brand joins the studio, whether we built it in-house or co-founded it with an outside operator. Four things get set up in the first 100 days. Roadmap, pricing, brand identity, and team deliberately don't change.
Co-founding a brand with an AI venture studio: what's actually different.
Domain scoping, equity structure, founder role, and day-to-day life all look different when you co-found a brand with a studio than when you raise a venture round and build alone. A practical guide for operators weighing the conversation.
NRR is the only B2B SaaS metric that compounds.
ARR is a snapshot. Net revenue retention is the engine. On a permanent-capital holding period, the gap between 105% and 125% NRR is the difference between a good business and a category-defining one.
SaaS roll-ups vs an AI venture studio: what actually differs, brand by brand.
Roll-ups bolt together companies that already exist. A venture studio builds each brand from a blank page and decides its outcome on its own merits. The operating posture, brand decisions, and life of the founding team are not the same.
AI doesn't lower SaaS prices. It widens margins.
The reflexive take is that AI commoditizes software. The actual outcome, for operators who run the math, is the opposite.
Operator continuity is the real asset.
Founder-led teams are the hardest asset to rebuild once they scatter. Patient capital and a studio structure are built to keep the operators who build a brand running it for as long as they want the job.
The holding period is the moat.
Pricing power, switching costs, distribution, talent — every B2B SaaS moat only fully compounds on a decade-plus clock. Time itself is the unfair advantage.
Permanent capital is the only sane structure for software.
Software compounds over decades. Most capital structures force exits in years. The mismatch is the entire opportunity.
// Subscribe
New essays, in your inbox.
One email when we publish. No drip campaigns, no upsells — just the work.