By clicking “Accept”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information, or manage your preferences.
Sunk cost may sink you.
For tutoring programs, networks, and edtech teams running their own virtual classroom — the math, the
migration, and the case for letting it go.
Written for the people who shipped it. Not the people who'll replace them.
Should we switch off our in-house virtual classroom to a commercial platform? If you built it before 2025 and have been operating it in production, almost certainly: yes.
Maintaining an in-house virtual classroom — whether it's a thin layer wrapping a CPaaS or a full-stack platform with your own SFU and native mobile — runs roughly $0.5M to $5M+/year, dominated by the loaded payroll of the small engineering team that has to keep it alive. The number on your spreadsheet is roughly your platform team's fully loaded comp, plus ~15% for infrastructure, compliance, hardware lab, and vendor stack. That's not a build cost; it's the annual cost of continuing. The build was rational. Continuing is a separate decision, made with a different formula and your own headcount in it.
Pencil Spaces handles the same surface area from $3,000/month on Scale (and from $6/month on the entry tier) — the full real-time stack, native mobile, compliance posture, and AI features. We've been operating it in production since 2020, with 10M+ sessions delivered across customers in 60+ countries, and we ship roughly two production releases a week, every week, for six years — a cadence in-house teams cannot match because their best engineers are maintaining the platform, not extending it.
And that's before the compliance and AI-attack surface that's compounded since 2020 — FTC enforcement against edtech (Edmodo, $6M, 2023), the largest breach of children's data in U.S. history (PowerSchool, ~62M students, December 2024), real-time deepfake video-call fraud (Arup, $25M, 2024). The platform owner is the auditee. The defense posture is what you're inheriting when you switch.
And to make the math impossible to refuse: Pencil Spaces offers The Switch Guarantee — migration is free, your Year-One cost is capped at what you're paying today to maintain your in-house platform ("Year One ≤ Year Zero," published in your contract), and we'll prioritize interviewing your platform engineers for roles on our team if their headcount becomes redundant. You never have a year on Pencil Spaces that costs more than the year before you switched.
Read the full case: the reframe · the annual-burn anchor · what you're maintaining · the cadence gap · the risk surface · the CFO frame · the promotionmigration plan · receipts · the Switch Guarantee · FAQ
We've been building a virtual classroom platform since 2020. We know what it took you to ship yours, because we shipped ours alongside you. The late nights with Safari's getUserMedia. The week your audio stack regressed on Samsung mid-range devices. The 4pm Eastern spike that exposed every concurrency bug at once. The SOC 2 evidence that had to age in time. None of this is in a marketing brochure for a reason.
So when we ask you to consider switching, we are not asking you to admit a mistake. Building was rational for the moment you built in. Pricing was different, commercial alternatives were thinner, AI hadn't bent the curve yet, and your program needed something the market didn't have.
The question is no longer "was it right to build?" Sunk cost is sunk; that question doesn't have a forward-looking answer. The question is "is it right to continue?" — and continuing is a different decision, with different math, made with different information than you had in 2020 or 2022.
Most of this page is that math. The rest is the migration story, because the math without a credible path forward is just a guilt trip.
The thesis, in one line. This page isn't an argument that building was wrong. It's an argument that maintaining is the next decision — and it's a different one.
What it costs to keep maintaining a virtual classroom you've already built — depending on what you built. The bottom of the range is a small team wrapping a CPaaS plus a UI layer; the top is a full-stack platform with native mobile, compliance, and AI. Your number is roughly your platform team's fully loaded payroll, plus ~15% for infrastructure, compliance, hardware lab, and vendor stack. Whatever number you land on, that's what commercial alternatives now compete with.
These are the clusters in a production virtual classroom — the same map your engineers know by heart. The point isn't intimidation; it's that none of these go away. Every one keeps an engineer awake on its own cycle.
It isn't a talent gap. It's a leverage gap. We've shipped roughly two production releases a week, every week, since 2020. Your team can't match that — not because they aren't capable, but because their best engineers are running the platform, not extending it.
Composite, drawn from a dozen real conversations with in-house teams between Jan and Apr 2026.
From our public changelog. Two releases a week. Six years running. The gap isn't widening because they got slower — it's widening because we never stopped.
Every week you spend maintaining the platform is a week your tutors don't get the AI features your students' parents are now expecting. Two releases a week, six years running — that's the cadence in-house teams cannot match, because their best engineers are answering pagers, not shipping the next thing.
Hey Jim,
Last Thursday a parent called the support desk saying their kid's tutor "sounded different" on Tuesday's session. We pulled the recording. The tutor sounded fine to me, fine to the parent eventually, fine to the platform team. But I spent that night reading through everything we'd have needed to actually verify — voice biometrics, deepfake detectors, session-anomaly flags, the chain of identity controls between the booking system and the live room. We don't have most of it. Some of it, we don't even have a vendor for.
I'm not writing because anything happened.I'm writing because I realized I wouldn't have known if it had.
You've been saying for two years that we should put the platform team back on the parts of the product that are actually ours, and that we should be on a commercial substrate. I gave you the same answer every time — that we'd built something good, that swapping the foundation felt risky, that the engineering pride was real. All of that's still true. None of it changes the math anymore.
You were right, Jim. I'm sorry it took me this long to see it.
I'd like to spend Friday with the team looking at Pencil Spaces and have a real number for you Monday — engineering hours, parallel-run plan, what we'd say to Aman, Ravi, and Sara.
Priya,
You don't have to apologize. The rest of us should have asked the question harder. I didn't.You held the position because you cared about the work, and the work was real.
Friday works. Block the morning. Bring the team you trust. I'll be there for the first hour and out of the way after.
Will do. One thing — I'd like to tell Aman, Ravi, and Sara before Friday.They deserve to hear it from me, and they deserve to hear what they're being moved to, not just what we're moving away from.
In May 2023, the FTC settled with edtech provider Edmodo for $6M over student data used in contextual advertising — the first FTC enforcement action specifically against an edtech vendor. The order required Edmodo to delete any models trained on student data collected without verifiable parental consent. The penalty was suspended on inability to pay; the order remains binding.
In December 2024, K–12 software vendor PowerSchool disclosed a breach affecting ~62M students and ~9.5M teachers across North America — publicly characterized as the largest breach of children's data in U.S. history. Class actions, AG investigations across multiple states, and follow-on extortion attempts against individual districts followed. The indemnification chain ends at the platform owner. Not your CPaaS vendor. Not your video-API supplier. You.
In January 2024, a finance worker at engineering firm Arup transferred $25.6M across 15 transactions after a video conference with deepfake renderings of the company's CFO and several colleagues — the entire call was synthetic, generated from publicly available executive footage. The same month, a deepfaked audio clip of a Maryland high-school principal went viral and required police forensics to clear his name.
For a learning platform specifically, the surface includes: tutor-impersonation deepfakes, voice cloning of staff for fraud, prompt injection of in-platform AI features (summaries, transcripts, coaching), adversarial inputs designed to elicit unsafe content for minors, and AI-generated CSAM as a new category of legal exposure. Most in-house teams scoped their security headcount for the 2020 threat model — not this one.
Commercial platforms operating at scale staff dedicated trust & safety, security engineering, ML-safety, and AI red-team functions. These aren't roles a maintenance team of 2–10 absorbs gracefully. They aren't roles you'd hire if you weren't already running the platform.
The asymmetry isn't talent — it's whether the specialization fits the job. Switching inherits our defense posture rather than asking your engineers to reproduce it. That's most of why a platform category exists at all.
Your platform team didn't sign up to be your trust-and-safety team, your privacy counsel's first call, or your AI red team. That's the part of the work that fits worst with what you hired them for. Inheriting the defense posture is what switching actually buys you — beyond the dollars and the cadence.
Sources: (proposed order filed May 22, 2023; ftc.gov); PowerSchool SIS incident disclosures (December 2024 – August 2025; powerschool.com); Hong Kong Police briefing on Arup deepfake fraud (Feb 2024) and CNN/Fortune reporting (May 2024); Baltimore County Public Schools / Baltimore PD investigation of the Pikesville High principal-deepfake case (January 2024). For attack categorization: NIST AI Risk Management Framework; OWASP Top 10 for LLM Applications.FTC v. Edmodo, LLC
Most CEOs of programs your size haven't run it. Their CFO has — and the gap between the in-house line and the commercial line is, increasingly, the conversation.
A fully loaded year for a small program wrapping a CPaaS plus a UI layer.Scale up by your team size.
One scenario inside a $0.5M–$5M+/yr range across the programs we talk to. A 2-engineer wrap of a CPaaS lands at the low end; a full-stack in-house build with native mobile and AI runs $4M+. The point isn't a specific number — it's the order-of-magnitude gap to the line below.
Real list pricing, no asterisks. Same SFU, whiteboard, recording, AI, mobile, compliance.
Custom-quoted at higher volumes. Even at 10x the base, the gap to in-house ongoing is at least an order of magnitude. Pricing is at .
pencilspaces.com/pricing
Continuing to pay your engineers to maintain a video calling platform when a better one exists, runs at a fraction of the cost, and ships AI features ten times faster — that isn't technical strategy. That's fiscal mismanagement. We say this respectfully, and to your CFO.
Your engineers are smart. They built a working virtual classroom from scratch. The question isn't whether they're capable. The question is whether the platform underneath is the work that deserves them.
WebRTC debugging. The Safari getUserMedia bug. The Firefox codec negotiation. The Chrome 121 release that silently changed MediaStreamTrack constraints.
Mobile binary maintenance. iOS review cycles. Android device fragmentation. Two app stores, two audio stacks, one perpetually broken Bluetooth path.
Compliance evidence. SOC 2 Type II controls that have to age in time. FERPA contract review. COPPA disclosures. Annual penetration tests.
The 4pm Eastern surge. Capacity planning, autoscale tuning, hot-row Postgres queues, pager rotations none of you wanted.
Recording pipeline ops. Compositor failures, encryption rotations, multi-region replication, retention policy enforcement, the silent failures nobody catches until parents email.
Trust & safety that grows with every district, every parent complaint, every consent flow change. Permanent overhead.
Tutor-matching algorithms tuned to your specific subject mix and learning-outcomes data — your moat, not anyone else's.
Curriculum delivery layer. The pedagogical surface that's the actual reason a parent picks you over a competitor. The thing your team has opinions about.
Outcomes data. Pre/post diagnostics, growth modeling, mastery tracking, the reporting layer your district contracts actually ask about.
Subject-specific UX. The math editor your competitors don't have. The reading-fluency assessment. The writing-feedback loop. The vertical craft.
Tutor-coaching features. Real-time feedback to your tutors mid-session. Pattern recognition across thousands of sessions. The thing that turns a tutor cohort into a flywheel.
Program operations. Scheduling that knows your business. Parent-communication flows that match your brand. Reporting your director actually opens.
Your team built something real. Let them build the next real thing.
Migration is the part that’s scared most teams off the decision — and the part we’ve gotten quietly very good at. Not a big-bang cutover. A parallel run, with a named technical lead from our team owning every edge.
Your technical lead and ours scope the data: recordings, session metadata, user records, schedules, custom fields. Export pipeline drafted in three days. No code changes on your side yet.
Pencil Spaces tenant provisioned, branded, integrated with your SSO and rostering. Both systems are now live. Your existing platform keeps running; new tutors can be onboarded to either.
Tutors move in cohorts of 5–15 with white-glove training. Students don't switch until their tutor switches. Sessions on the new platform run side-by-side with sessions on the old.
By the time you cut over, every active session has been on Pencil Spaces for weeks. The cutover is administrative, not operational. Old platform stays in read-only mode for archive access.
All of it is exportable and importable. Recordings re-host into our storage with their original timestamps preserved. Session metadata, user records, schedule history, attendance, tutor assignments, custom fields — we read your schema; you keep the source.
Nothing is left behind unless you ask us to leave it. Old platform stays in read-only mode for as long as you need.
Not if it's parallel. Tutors are onboarded in cohorts with named training sessions; students don't switch until their tutor switches. By the time you cut over, every active session has already been running on Pencil Spaces for weeks.
We've onboarded tutor populations from a dozen to a thousand. The training material is tested.
The honest answer: this is the question we take most seriously, and it's where we work hardest. Switching is a promotion for that team, not a layoff. The engineers who built your in-house platform are exactly the engineers your differentiated product needs.
We work with engineering leaders directly during transition: documented exits, clear next-mission statements, and time to ramp.
The numbers we hold ourselves to. 2–4 weeks for small programs (under 1,000 user-hours/month). 4–8 weeks for mid-size programs. 8–16 weeks for complex networks with custom integrations or strict procurement. We've moved programs of varying sizes off in-house systems and have not had a cutover go badly. Your migration plan is built around your network, your district contracts, and your tutor calendar — not ours.
These aren't projections. They're the operating envelope of the platform you can buy from us today — running in production, every day, since 2020.
I've been the founder on the other side of this conversation. In 2020 my co-founder Amogh and I walked away from senior tech-leadership roles at companies like Meta and Google to make the bet on building a real-time, virtual-classroom-grade platform from first principles. The decision wasn't wrong — we had a thesis about what tutoring needed that nothing on the market did. The cost — in years, in headcount, in the long tail of work nobody warned us about — was real.
The thing nobody told us — and the thing I now tell every founder who asks — is that the infrastructure isn't the product. It's the part of the product you're forced to ship before you get to ship the part you actually care about.
Six years later, that infrastructure is what we sell. As a full-stack platform, called Pencil Spaces. As an embeddable API, called Carbon. We built it once, at unsustainable cost, because we had to.You don't.
And if you're reading this page, you've already done version of what we did. The question isn't whether you have the talent. You did. You shipped it. The question is whether you want them on the platform, or on the part of the product nobody else in your category has the right to build.your
We're not the only company that's tried to build a virtual classroom from scratch. Two of the most prominent venture-backed attempts of the last five years, and what they had to raise:
$33M led by , et al.Series A:Maveron, Corner Ventures
Founded by veteran ed-tech operators with deep institutional backing. Focused on higher-ed and enterprise; built much of the real-time stack from first principles.
$105M led by (July 2021).Series B:SoftBank Vision Fund 2
Built virtual-classroom UX on top of Zoom's Meeting APIs — outsourced the hardest layer of the stack to a third party, and still had to raise nine figures to ship around it.
Read that second card carefully. Class raised more than five times the cost of building a virtual classroom from scratch — and didn't even build their own real-time video. They wrapped Zoom and built UX on top. Even that was a $160M+ undertaking. Engageli, building closer to the metal, raised $47.5M and runs leaner. Neither story is a knock on the founders — both companies have credible operators. It's a knock on the assumption that this category is cheap to keep running. It isn't. If they couldn't justify continuing the build, ask whether you can justify continuing to maintain yours.
Sources: public funding announcements, Crunchbase, Tracxn, PRNewswire (Class Series B, July 2021). Verify before relying.
Most of what makes a tutoring program win has nothing to do with whether the video feed renders on Safari 14. It has to do with the curriculum, the tutors, the matching, the outcomes data, and the parent experience. Those are the things only your team can build, and the things your competitors are also trying to build — with their full engineering attention pointed at them.
Continuing to maintain an in-house platform doesn't just cost dollars. It costs the years of compounding focus those engineers could be putting into your moat. Every quarter spent on the substrate is a quarter your competitor spends on the surface.
We will handle the video, the whiteboard, the recording pipeline, the mobile binaries, the compliance evidence, the on-call. We've already done it.You don't have to do it again.
Four commitments to programs migrating from an in-house virtual classroom. We make them because the math works for both sides — you exit a recurring engineering line that should never have been a recurring engineering line, and we add a customer that becomes a multi-year revenue line at our marginal cost. We are betting on Year Two and beyond.Published. In your contract. Signed.
Named technical lead from our engineering team. Four-week parallel run with both platforms live. Full transfer of users, recordings, integrations, and data. Zero implementation fee. We absorb the migration engineering cost on our side because we're the ones with the leverage to do it cheaply.
For most workloads, Pencil Spaces Scale costs less than a single fully-loaded senior engineer. Eliminate one role on the team that maintains your in-house platform and the platform line is already net-positive in Year One — before any other savings show up.
Your Year-One Pencil Spaces cost is capped at what you're paying today to maintain your in-house platform — engineering payroll, infrastructure, compliance, vendor stack, the whole line. From Year Two, the line goes down. You never have a year on Pencil Spaces that costs more than the year before you switched. Published, in writing, in your contract.
If your maintenance headcount is going to be redundant, we prioritize interviewing those engineers for engineering roles on Pencil Spaces. Their domain expertise is real and rare. We'd rather have them working on this problem at scale than have you carry severance.
Not a demo. Not a trial. A conversation about your specific situation — your stack, your network, your tutor population, your district contracts — and what the four commitments look like in your contract.
30 minutes. No demo deck. Bring your CTO and your CFO. We'll bring the migration plan and the term sheet.
Maybe. The honest answer to most of these is: probably less different than you think. Here are the questions we get most often when this conversation starts.
