8 bits for a Byte: Block just cut 40% of its workforce – not because business was bad, but because it believes AI already works well enough to run the company at half the headcount – a bet built on years of headcount caps, deliberate de-duplication, and automation already handling 99.95% of transactions – most companies talk about AI-native operating models; Block filed the paperwork and handed out 20-week severance checks – the cultural and organizational prerequisites that made this possible took years to build, and most firms don't have them – this week, we connect every dot.
I want to be direct about where I stand. I don't believe AI reduces employment in the long run. I believe it flattens it. The workforce doesn't shrink; it redistributes — toward the long tail of smaller businesses, the ones born compounding, the ones that never needed a $450M restructuring charge to right-size their architecture. Enterprise organizations will shed the layers of process and coordination complexity that accumulated over a century of industrial-era management — not because they want to, but because AI agents can absorb and simplify those layers faster than any org chart redesign ever could. The headcount that leaves the enterprise doesn't disappear. It reappears in a thousand companies like Every.to, building leaner and compounding harder from day one. Block's announcement is not a signal that work is ending. It's a signal that work is relocating.
Tell me what landed – the quick survey at the end directly shapes what I create next.
Let’s Get To It!
Welcome To AI Quick Bytes!
Pneumonia has a way of clearing your calendar and your thinking simultaneously. While I was down this past month, Block made what I believe is the most important organizational announcement of 2026. So this week, we go deep — not just on what Block did, but on why it signals the beginning of a structural reordering that most enterprise leaders aren't ready for, and what a 20-person company called Every.to reveals about where this is all heading.
Welcome To AI Quick Bytes!
Bit 1: WHAT BLOCK ACTUALLY DID (AND WHAT THEY DIDN'T SAY)
On February 26, 2026, Block announced a reduction of more than 40% of its workforce — from just over 10,000 employees to just under 6,000. Estimated restructuring charges: $450–$500 million. Execution timeline: substantially complete by end of Q2 2026. The shareholder letter was emphatic that 2025 performance was strong with Gross Profit of $10.36 billion Operation Income of $2.08 billion.
Read that again. Block didn't cut because revenue collapsed. It cut because leadership believes "intelligence tools" have permanently changed what it means to build and run a company. That is a different argument than "macro headwinds" — and it carries a different weight. Jack Dorsey's memo described a deliberate choice to execute a single deep cut rather than repeated rounds, framing staged reductions as "destructive to morale." He's right about that. Every leader who has lived through three rounds of layoffs in 18 months knows what happens to the culture in between — people stop investing and start optimizing for survival. The severance package — 20 weeks of salary, one additional week per year of tenure, six months of health care, and $5,000 in transition support — signals a leadership team that had the financial capacity to protect departing employees rather than just protect the balance sheet. I have deep respect for that choice. One clean cut, done with dignity. That is what accountable leadership looks like.
Bit: Block didn't cut because it failed; it cut because it believes it already won the automation argument.
Three Key Takeaways:
Block's gross profit per employee is projected to nearly double — from ~$1.02M at end of 2025 to ~$2.03M post-reduction — using existing gross profit guidance and a 6,000-person headcount floor.
The $9.2 billion in total liquidity disclosed at year-end made a $450–$500M one-time charge absorbable without triggering a cash crisis; most companies don't have that runway.
Framing matters: "operating-model redesign" rather than "cost reduction" is not spin — it reflects a genuine architectural commitment with measurable milestones tied to gross profit per employee.
ACTION BYTE: Map your own gross profit per employee trend over the last three years and bring the chart to your next leadership offsite — if the line is flat while headcount grew, you are already having Block's conversation, whether you know it or not.
Bit 2: CENTRALIZATION AND FLATTENING AS EXPLICIT DESIGN CHOICES
Block's annual reports and Dorsey's memo converge on a phrase that is easy to overlook: "smaller and flatter teams." These are not synonyms. Smaller reduces headcount. Flatter removes managerial layers and shortens the distance between a decision and its execution. Block is doing both simultaneously — and each enables the other in a reinforcing loop.
The GE of the 1990s demonstrated that radical delayering accelerates decision speed — Jack Welch removed multiple management layers across a 300,000-person organization and watched cycle times compress. Block is operating at a fraction of that scale, but the principle holds: every management layer that exists to translate strategy into execution is a layer AI-assisted coordination can now absorb. Block's human capital disclosures explicitly set manager expectations to "break down silos and hierarchy." That is not cultural aspiration language — it is a structural directive backed by performance review criteria. Speed is no longer a side effect of good culture; Block has made it a measurable organizational requirement. Every.to's Dan Shipper put the same principle more directly: two-pizza teams are now themselves too big. His "two-slice team" model — one person per product — is the logical end state of the architecture Block is building toward. Block is paying to get there. Shipper built toward it from the first commit.
Bit: Flatter isn't just cheaper; it's faster — and speed compounds when the architecture underneath it does too.
Three Key Takeaways:
Centralization removes duplicated functions across product lines — but it requires strong shared standards and tooling to prevent service degradation; Block has been building both for years.
Flattening demands clearer individual accountability — when layers disappear, each remaining person owns more surface area, which only works if the tooling reduces cognitive load proportionally.
Block explicitly frames "speed" as a structural expectation, not a cultural aspiration — that distinction signals a performance management shift, not a motivational poster.
ACTION BYTE: Count the management layers between your CEO and the person writing customer-facing code or copy — if the number exceeds four, you have structural latency that no AI tool will fix until the layers themselves are addressed.
Bit 3: THE MULTI-YEAR GROUNDWORK MOST COMPANIES SKIPPED
Block's 40% cut did not appear overnight. In November 2023, the company disclosed an absolute headcount cap of 12,000 employees — and committed to operating below it through performance management, scope prioritization, and explicit centralization to reduce duplication. By end of 2024, headcount was already at 11,372. By end of 2025, it was 10,205. The culture had already normalized the idea that headcount is a constrained resource, not an appetite.
Think of this as the organizational equivalent of the Japanese manufacturing concept of muda — eliminating waste before a crisis forces you to. Block's leadership spent two years telling managers to trade scope for staffing, consolidate overlapping teams, and treat duplication as a defect. When the 2026 cut arrived, it wasn't the first conversation about organizational efficiency — it was the final one. That sequencing is what made a 40% reduction feasible rather than catastrophic. You cannot skip this step. Companies that announce copycat cuts without this groundwork will remove connective tissue and leave the redundant structures that created the problem in the first place.
Bit: The cut wasn't the strategy; years of deliberate constraint were.
Three Key Takeaways:
- Operating under a publicly disclosed headcount cap changes manager behavior — it forces trade-off discipline that most companies only attempt during downturns, not in advance.
- Block's prior cost-efficiency actions already produced $78.6 million in severance-related expenses in 2025 and $26.8 million in 2024 — evidence of ongoing restructuring practice, not a one-off shock.
- Companies that announce large AI-driven cuts without prior de-duplication work risk removing "connective tissue" while leaving redundant structures intact — the worst of both outcomes.
Action Summary: Begin the groundwork this quarter.
1. Audit team charters: identify every function duplicated across more than one product line or business unit this quarter.
2. Set a soft headcount cap by department; communicate it to managers as a design constraint, not a threat.
3. Track scope-to-staffing ratios quarterly — if scope shrinks but headcount holds, the ratio is broken.
4. Centralize one duplicated function (analytics, tooling, or risk) as a pilot before year-end.
5. Schedule a leadership review in 90 days to assess whether the pilot reduced coordination overhead or created new bottlenecks.
Bit 4: THE ARCHITECTURE THAT COMPOUNDS — WHAT BLOCK IS BUILDING AND EVERY.TO ALREADY RUNS
Block's operating model shift hinges on a specific architectural choice: moving from "teams that build features" to "teams that build reusable capabilities." The shareholder letter describes a future where customers build their own features directly on top of Block's capabilities through composable interfaces, with an intelligence layer orchestrating operations underneath. The four focus areas — customer capabilities, composable interfaces, proactive intelligence, and an intelligence model to fully orchestrate operations — are not product categories. They are organizational load-bearing walls.
This is the same design logic that drove Amazon's 2002 "API mandate" — the internal directive that every team must expose its capabilities through a service interface, with no exceptions. That decision made AWS possible. Block is making the analogous bet: if the capability layer becomes the shared contract between teams and customers, large coordination layers disappear because the architecture carries the weight they once carried. Now consider what that architecture looks like when you build it from the first line of code. Every.to — a media and software company running six business units and four software products on 20 full-time employees — calls this compound engineering. Kieran Klaassen, GM of their email product Cora, formalized the principle: each unit of engineering work should make subsequent units easier, not harder. Bug fixes eliminate entire categories of future bugs. Patterns, once codified, become tools for the next build. The result is a codebase that compounds capability rather than compounds debt. Naveen Naidu maintains a 143,000-line production codebase — Monologue, used 30,000 times daily, transcribing 1.5 million words — solo, with AI. Block is paying $450–$500 million in restructuring charges to reach a comparable architecture. Every.to started there.
Bit: Capability platforms don't just serve customers; they compound — and the companies born compounding never pay the conversion cost.
Three Key Takeaways:
- "Composable interfaces" reduce the need for custom implementation work — each new customer use case draws from existing primitives rather than spawning a new project team.
Compound engineering's core unlock: when bug fixes eliminate categories of future bugs, you don't just fix problems faster — you reduce the total stock of future problems, permanently.
- The architecture and the headcount model are inseparable — you cannot cut 40% of your people without first building the technical scaffolding that replaces their coordination work, and that scaffolding must compound, not accumulate.
Action Summary: Audit whether your organization compounds or accumulates.
1. Ask your engineering lead: does a bug fix eliminate a category of future bugs, or just the current one? The answer tells you whether your codebase is compounding capability or accumulating debt.
2. Map every team that exists primarily to coordinate between two other teams — those are the first candidates for automation or architectural elimination.
3. Identify which product lines expose reusable capabilities versus bespoke builds; close that gap over 12 months.
4. Require product and engineering leads to present a "capability reuse ratio" in quarterly business reviews — this is your compound rate.
Bit 5:
Quote of the Week:
The companies that will lead the next decade are not the ones restructuring into AI-native — they are the ones that built compounding capability from the first line of code, so the restructuring conversation never had to happen.
Bit 6: Sunday Funnies
Bit 7: WHY MOST COMPANIES CANNOT REPLICATE THIS NOW
Amazon cut roughly 30,000 corporate roles across October 2025 and January 2026 — approximately 10% of its corporate workforce — citing bureaucracy reduction and AI-driven efficiency. Pinterest cut nearly 15% in January 2026. eBay cut 6% on the same day Block announced its reduction. All cited AI. None approached 40%. The differentiator is not that Block is using AI while others aren't. The differentiator is that Block is proposing a company-wide operating model built around AI-mediated orchestration — and cutting deeply as if that model is already validated internally.
Three constraints explain why most firms are watching rather than following. First, the proof threshold: Block proceeds despite its own disclosure that AI reliance may not deliver intended benefits. Most organizations need demonstrable ROI before accepting that execution risk. Second, the complexity prerequisite: without years of de-duplication work, a 40% cut removes connective tissue while leaving redundant structures intact. Third, the liquidity requirement: with $9.2 billion in total liquidity against a $450–$500 million charge, Block could absorb the hit without a balance sheet crisis. But here is what none of these three addresses: compounding readiness. Here is my honest read of where this ends for the Fortune 500 over the next decade — most won't make this transition successfully. Not because AI won't force the issue. It will. But the AI-native competitors are already here, and the talent drain will accelerate everything. The engineers who can run what Naveen Naidu runs at Every.to are not waiting for a large company to restructure. They are already building companies that compound from day one. The enterprise talent moat doesn't drain all at once. It drains one resignation letter / one restructure at a time.
Bit: AI readiness isn't about tools; it's about whether your architecture compounds capability or accumulates debt.
Three Key Takeaways:
Scale of cut (40%+) versus peers (6–15%) reflects a genuinely different operating model thesis — not a more aggressive cost-cutting mandate.
The single-round decision is itself a cultural artifact: only a leadership team managing to a headcount cap for years can credibly execute "one clean cut is better than repeated reductions."
Reuters commentary raises a fair counterpoint: AI can serve as convenient cover for slimming organizations that over-hired post-pandemic — evaluating both explanations honestly is part of your strategic due diligence.
Action Summary: Use Block's framework as a readiness audit — and add the compounding test.
1. Score your organization on four dimensions: automation baseline (1–10), de-duplication progress (1–10), leadership risk tolerance (1–10), and compounding architecture (1–10) — a combined score below 28 means you are not ready for a Block-scale move.
2. Identify your binding constraint and build a 12-month plan to close the gap.
3. Pressure-test your "AI efficiency" narrative with your CFO — can you quantify what automation is already saving, in real dollars, before you cite it as a restructuring rationale?
4. Review your liquidity position against a hypothetical 15–20% restructuring charge — know your actual financial tolerance before setting a scale target.
Bit 8: WHAT YOU SHOULD DO THIS WEEK
Block's moves are instructive not because you should replicate them — but because they reveal the minimum prerequisites for an AI-native operating model to become a genuine organizational option. The companies that will have this conversation on their own terms in 2027 and 2028 are doing the structural groundwork today. The ones still waiting for AI to "prove itself" before changing anything are building debt, not readiness.
The historical lesson is consistent. When cloud computing shifted from experiment to infrastructure between 2008 and 2014, organizations that had already centralized IT procurement and standardized their dev environments made the transition in 18 months. Those with fragmented, duplicated, siloed infrastructure spent five years rationalizing the mess before they could accelerate. Block's playbook is the cloud transition applied to the organizational layer. But here is what should keep you up at night: Every.to isn't doing the cloud transition. They were born in the cloud, building on compound engineering principles from the first commit. Dan Shipper built Proof — an agent-native markdown editor with full collaboration and AI provenance tracking — in his spare time. That would have required three to four engineers and six months, previously. Every’s Kieran Klaassen's compound engineering philosophy has 7,000 stars on GitHub because builders everywhere recognize what it represents: the operating system of the 21st-century company. Not a technique. An architectural conviction that every team, every workflow, and every product decision should compound the next one. Block is paying $450–$500 million to reach this architecture. You don't have to. The philosophy is documented. The tools exist. The only thing required is the organizational conviction to build this way before the market forces you to.
Bit: Optionality isn't built during the disruption; it's built by compounding capability in the years before the disruption arrives.
Three Key Takeaways:
The question isn't whether to build an AI-native operating model — it's whether you are building a compounding one, so the option is available and affordable when you need it.
Compound engineering is not an engineering philosophy — it is an organizational one: the principle that each unit of work should make subsequent units easier applies to teams, processes, and architecture equally.
Block is the leading edge of enterprise transformation — but the true leading edge is the organization that never needed to transform, because it built compounding capability from the first commit.
Action Summary: Start the groundwork this quarter.
1. Assign a senior leader to own a "structural readiness" audit across automation, de-duplication, organizational flatness, and compounding architecture — due in 60 days.
2. Run a headcount cap exercise: if you had to operate 20% leaner tomorrow, what would break and what would be fine? The answer tells you more than any consultant's assessment.
3. Read Kieran Klaassen's compound engineering guide at every.to — then ask your engineering and product leads whether your architecture passes the compound test.
4. Add "gross profit per employee" to your monthly leadership dashboard alongside revenue per employee — watch both lines.
5. Decide on one function to centralize as a proof-of-concept this year; measure coordination overhead before and after, and ask whether the centralized version compounds or just consolidates.
Until next time, take it one bit at a time!
Rob
P.S. Thanks for making it to the end—because this is where the future reveals itself.
This Week's Resource:
I've referenced Every.to throughout this edition — and I want to be direct: I recommend it because I believe in what they're building, not because it's convenient to the story.
Every.to publishes the sharpest thinking and implementation guides on AI-native business models available anywhere. If the compound engineering philosophy, the two-slice team model, and the question of "what does a company built to never restructure actually look like?" are relevant to your work — and if you're reading this newsletter, they are — Every.to is where that conversation lives at the highest level.
Start reading here: every.to
Forward this newsletter to the one person on your leadership team who should be thinking about this. They'll thank you.
Full transparency: above links to Every are referral links — I'd send you there anyway, but I'm not above getting paid to point you toward the best thinking on the internet :-).