long-form-interview· Dick Costolo

Twitter Ex-CEO Dick Costolo: The Operator's Case Against Founder Mode

Costolo''s decision-velocity stack at scale: (1) Push decisions down the stack — DRIs, not committees. (2) Bias-to-yes — only your direct manager or legal can say no; no fiefdoms can require approval. (3) Never invoke "so-and-so said" — context, not authority. (4) Don''t solve problems with processes — solve them with DRIs and operating control. (5) Manage outcomes, not processes — leadership''s job is correcting mistakes quickly, not preventing them.

twittercostolofounder-mode-countervelocitybias-to-yesdriinstagram-regretsteve-jobshalligan95% confidence

Why this is in the corpus

Strongest cross-corpus complement to founder-mode pattern — Costolo offers the operator''s case for non-founder professionalism done well. Adds the Steve Jobs Pixar information-gathering meeting play, the bias-to-yes enforcement rule, and the trans-fire block to the corpus. Sharpens "fire fast" pattern with the manage-out-low-performers discipline.

Summary for skimmers

Joined Twitter as COO when company was a collective making decisions by group vote. First fix: velocity. Pushed decisions down the stack, killed committees, banned "so-and-so said" reasoning. Bias-to-yes: only direct manager or legal can say no. Solved problems with DRIs, not processes. Make sure everybody understands what you understand. Used Steve Jobs Pixar method — direct one-on-ones with ICs without leader present. Trans-fire block prevented underperformers from moving teams. Regret: not buying Instagram for $700M before Facebook did. Lesson: don''t copy the archetype book — nobody is fooled.

Briefing

What survives the editorial filter

This page should feel like a smart colleague already listened for you and left only the operating logic worth keeping. Not everything said in the episode makes it through.

Trust signal

direct_practitioner_account

Guest type: practitioner.

Best used for

Decision-grade retrieval metadata not yet added for this episode.

Hold lightly

No explicit downgrade reason stored yet for this episode.

Principles

Durable claims that survive beyond the speaker's biography — each with explicit limits, transferability judgment, and evidence.

Principle

Aggressive ≠ jerk — direct + forthright with empathy in the relationship, not in the message

Direct-but-empathic feedback compresses delivery time and preserves the relationship; fluff-dressed feedback extends pain and dilutes the signal.

For your next tough-feedback conversation: write it down the night before. Say only what''s on the page. Then sit in silence. Empathy goes in the relationship, not in the message.

I'm super impatient. ... I'm just direct and blunt with people. I'm not a jerk. ... a lot of times, especially engineering leaders ... they think when you tell them you have to manage poor performers out, they think you have to be a jerk when you do that. You don't have to be a jerk about it. You just have to be about direct.Dick Costolo
Bill Campbell actually gave me this advice early on — is, like, write down what you want to say, and then say that and don't say anything else.Dick Costolo

Principle

Bezos rule: only take long on decisions that are existential

Decisions split into one-way doors (existential, hard to reverse) and two-way doors (reversible) — applying existential-grade deliberation to two-way doors wastes velocity that compounds.

Classify your 5 active major decisions: one-way or two-way? Apply existential deliberation only to one-way doors.

This is the old Jeff Bezos thing. Only take a long time to make decisions that are existential. ... There are fewer and fewer of those decisions that can be like, if we screw it up, we can go back and do the other one. Do those right away.Dick Costolo

Principle

Forestry management beats forest-fire fighting at scale

Reactive firefighting scales sublinearly (one leader can only fight so many fires); strategic forestry management scales because it changes the conditions under which fires happen.

Track how much of your week is fire-fighting vs forestry. If <30% is forestry, you''re still in tactics.

Your job as a director, as you grow in this company is you can't just say, Well, I put out this forest fire. ... your job is to map the territory and start to do forestry management. ... You need to be focused more on thinking about this whole pie and what has to be true 24 months from now so this whole territory is in better shape.Dick Costolo

Principle

Make sure everybody understands what you understand — priorities + context + personal stakes

Context-rich priority communication produces compound motivation (the bricklayer building a cathedral); priority-only communication produces compliance but not initiative.

For your top 3 priorities, ask each direct report: do they know what success means for them personally if we hit it? If not, you''ve communicated the what but not the why-for-them.

Ben had in his outline ... make sure everybody understands what you understand. So I would always tell my managers that. Like, your job is to make sure everybody understands what the priorities are and what it means for them and their team, how they fit into those priorities, and more importantly, for your team, what it means for them personally and professionally if we achieve what we have to achieve.Dick Costolo
It's like the analogy of somebody asking a bricklayer what he's doing, and the bricklayer says, "Oh, I'm laying bricks." And the other bricklayer says, "I'm building a beautiful cathedral."Brian Halligan

Principle

Push decisions down the stack — kill committees, ban "so-and-so said"

Group-decision-by-committee makes decision speed scale O(n) with group size; pushing decisions down the stack with banned authority-citation makes it O(1) at the level where the work happens.

Audit how many decisions came up the stack this week that should have stayed down. Each one is a vote that you''ve created a committee-by-proxy.

We're not going to make decisions as a group, we're going to push decisions down the stack. ... don't come to me and ask me what you should do about this unless you have no idea. ... don't tell your team what to do. Tell them to figure it out and push as many decisions down the stack as you can.Dick Costolo
God, I hate hearing "Dick said" in the company as a reason for things. ... part of the speed was no, we're not going to have so-and-so said. Push the decisions down the stack.Dick Costolo

Principle

Solve problems with DRIs, not processes

Adding process steps to prevent specific failures accumulates monotonically — each failure adds a step, no failure removes one, so processes grow unboundedly while outcomes stagnate. DRI ownership inverts the math.

For every process step in your org, ask: is there a DRI who owns this outcome and could replace this step? If yes, replace it.

Don't solve problems with processes. ... at Google famously, the launch checklist got to be, like, 17 pages long. ... pretty soon you're not managing to outcomes, you're managing to the processes.Dick Costolo
You should probably only really have a couple organizational processes, and everything else should be span of control and decision-making. ... It's not the job of leadership to prevent mistakes from happening. It's the job of leadership to correct mistakes quickly when they happen.Dick Costolo

Principle

There are many ways to climb the mountain — don''t copy the archetype

Archetype-imitation fails twice: (1) the source biography is incomplete, so the imitator is copying a fiction; (2) the imitator doesn''t hold the underlying conditions, so the imitation looks performative and the team isn''t fooled.

For every imitation move you''re making — Jobs-style, Musk-style — ask: is this who I actually am? If no, the team can already see the gap.

I go back to the Jeff Bezos comment to Jack Dorsey. "Jack, there are lots of different ways to be successful." I think the mistake that CEOs make is they read the Walter Isaacson book on Elon Musk or the one on Steve Jobs and they think this is the map.Dick Costolo
If you try to be someone you're not, nobody is fooled. ... Nothing is more miserable than trying to be someone you're not.Dick Costolo

Frameworks

Reusable systems and operating models — including when they help and when they break.

Framework

Framework: Forest fires vs forestry management

The director who keeps proudly putting out fires has failed the strategic-leadership test.

You can only work so many hours/day. Stamping fires uses 100% of time and produces no compounding effect. Forestry asks "what has to be true 24 months from now."

Your job as a director is to map the territory and start to do forestry management. Stamping out fires isn''t going to work as we scale.Dick Costolo

Durability: Durable. The "tactical-fire-fighter cannot scale" pattern is structural.

Named mental model with clear diagnostic.

Framework

Framework: Make sure everybody understands what you understand (Horowitz)

Communication architecture is the most under-invested system in scaling companies.

Employee who understands "why" can self-correct in novel situations. Employee who only knows "what" needs instruction for every variation.

Make sure everybody understands what you understand. Your job is to make sure everybody understands the priorities and what it means for them personally and professionally if we achieve what we have to achieve.Dick Costolo (citing Ben Horowitz)

Durability: Durable. The "context-sharing produces self-direction" mechanism is structural.

Meta-framework for delegation.

Framework

Framework: Bias to Yes — only your manager can say no

Eliminates the silent-no failure mode where you''re not told "no" but routed through 14 approvers who must each say yes.

Standard org-design generates implicit veto rights at every functional boundary. Each veto-holder rarely says no — they say "go ask these 14 people." Compounds into months of paralysis.

Bias to yes is: only the person you report to is allowed to tell you you''re not allowed to do that. Or legal can tell you if you do that you''ll be fired. But no other organization is allowed to tell you you can''t do that.Dick Costolo

Durability: Durable. The "silent no via approval chain" pattern is structural to org growth.

Most actionable Costolo idea.

Signals

What appears to be shifting, for whom it matters, and what happens if you ignore it.

Signal

Signal: AI two-way doors are reshaping the focus-vs-breadth trade-off

Cost-of-reversal has dropped, raising optimal parallelism in CEO time allocation.

AI-accelerated development collapses build time from quarters to weeks. Failures caught faster, recoveries cheaper. Bezos''s do-everything stance applies more broadly.

Focus is never overrated. However the fact that you can do more quickly and these are two-way doors. There are fewer and fewer existential decisions.Dick Costolo

Durability: Time-sensitive. The "two-way door expansion" is current-era specific.

Forward signal.

Signal

Signal: Speed/velocity is now the universal CEO obsession

Velocity is the de facto first-order CEO metric in 2026.

AI-era markets reward shipping speed. Slow companies lose to fast ones. CEOs who preserve early-stage velocity through scale capture disproportionate value.

All any CEO wants to talk about now is speed. You get bigger and get less done. Barnacles build up.Brian Halligan (confirmed by Costolo)

Durability: Time-sensitive on the cultural shift; durable on the underlying advantage.

Two CEOs confirming.

Opportunities

Only included where there is a buyer, a real wedge, and a plausible revenue path — not vague idea theater.

Opportunity

Opportunity: Tooling for distributed management-by-walking-around

Tools that surface ship-rate-by-individual + automatic CEO-recognition prompts integrated with Slack, GitHub, Linear.

GitHub + Linear + Slack data can surface "who shipped what recently." Product is CEO-facing dashboard: "Marcel shipped retweet-v2 this week, here are 3 metrics improvements."

Bill said you could roll a grenade into Twitter at 5:30 pm. So I''d come back at 9-9:30 pm and whoever''s here, I''d talk about their work at the next all-hands. Slowly the culture changed.Dick Costolo

Durability: Time-sensitive. 18-24 month window before Linear/GitLab build it in.

Specific market gap.

Lessons still worth keeping

Useful takeaways that did not fully clear the bar for durable principle status.

Lesson

Lesson: The "Dick said remove the JSON" rumor

False CEO-attribution is the universal cover for unpopular technical decisions.

When ICs push back on a manager''s call, the easiest defense is "the CEO said." Terminates discussion, shifts blame upward. At scale, every difficult decision becomes "Dick said."

This Android engineer asked: did you say remove the JSON file? I was like, I know what a JSON file is but I couldn''t tell you what''s in the Android boot JSON. Some manager decides and the engineer pushes back, and they go: Dick said do it.Dick Costolo

Durability: Durable. The false-attribution-as-cover pattern is structural to org hierarchy.

Specific named incident.

Lesson

Lesson: The dirty-dishes "this is why the site crashes" email

The CEO''s job in a low-ownership culture is to find the smallest visible symbol of the problem and address it with maximum visibility.

Dishes are a permanent reminder of "we tolerate this." Connecting the visible-permanent problem to the invisible-intermittent problem makes the abstract concrete.

The kitchen was just a mess. Dirty dishes in the sink. I sent out an email: this is why the fucking site crashes. Because everything''s ''oh well, someone else will clean it up.''Dick Costolo

Durability: Durable. The small-symbol pattern is structural to behavioral change.

Memorable incident with transferable lesson.

Lesson

Lesson: Bezos "I like to do everything" vs Jobs "say no to almost everything"

Two of the highest-leverage CEOs operated from opposite operating philosophies. Focus is a tool, not a moral law.

Jobs''s focus was right for Apple — capital-intensive hardware where one wrong bet sinks the company. Bezos''s do-everything was right for Amazon — two-way doors where parallel experimentation surfaces winners cheaply. Doctrine matches cost-structure of mistakes.

Steve says the CEO''s job is focus, say no to almost everything. Bezos looks at Jack and goes: ''Oh yeah? Well, I like to do everything.''Dick Costolo (recounting Bezos)

Durability: Durable. The "cost-of-mistake determines doctrine" principle is structural.

Two named operator philosophies in direct conflict.

The Plays

Try these this week

Verb-first executable actions — each one tied to a stated outcome in the episode.

Steve Jobs Pixar information-gathering meeting — sit with team without leader present

Outcome: Direct-with-IC meetings without the leader present surface signal the leadership-chain filters out, while the "I''ll bring this back to your leader" framing preserves chain of command.

When Steve was managing both Pixar and Apple ... he'd go into the office on Friday, and he would say, "I want to sit down with the illustrators, without Lasseter, the director of the movie." And he would go — there are 8 or 9 of them — "Janet, tell me something that's not working well at Pixar." ... At the end of the meeting ... "Okay, here's what I heard. I'm going to go talk to John about it and he'll deal with the rest on Monday."
Dick Costolo
45-60 minutes per session; rotate teams over the quarter per (proposed)
  1. 1

    Schedule a 45-60 min meeting with 8-9 ICs from one team

    Without their leader. Pick a team where you want to understand priority-alignment.

  2. 2

    Ask each person: "Tell me something not working well"

    Then: "Tell me something working well." Not about their job — about the team and priorities.

  3. 3

    Take notes; don't agree or disagree

    "Bob, do you agree with that?" Sample across the room. You're mapping the view, not debating it.

  4. 4

    Close with the explicit feedback path

    "Here's what I heard. I'm going to go talk to [leader] about it Monday — they'll follow up with you."

  5. 5

    Bring the signal back to the leader on Monday

    Chain of command stays intact. If the leader hates this technique, that's a strong leadership-signal worth investigating.

Stop or pivot when

  • Leader is not in the room
  • CEO closes with "I'll talk to your leader" — no end-runs
  • Hostility from the leader is a signal worth investigating

Scripts

Before you start

  • · Org structure with intermediate leaders (CEO not directly managing ICs)
  • · Discipline to bring signal back through the chain, not around it
  • · Willingness to interpret leader hostility as data, not pushback
operating-cadenceorganisational-designexecutive-hireseries-bseries-cgrowth-stagepublic-companylate-stage

Play: Steve-Jobs-style listen-without-leadership team meeting

Outcome: Diagnostic for whether the team understands priorities (and whether manager has been honest).

Context: When manager is in the room, ICs filter their answers. Without the manager, ICs speak more candidly. The diagnostic isn''t the specific complaint — it''s whether IC version of priorities matches management version.

Steve would say: today I want to sit down with the illustrators, without Lasseter in the room. Janet, tell me something that''s not working at Pixar. At the end he''d say: here''s what I heard. I''m going to talk to John about it on Monday.
Dick Costolo (citing Steve Jobs)
60 min per session + 30 min manager debrief per
  1. 1

  2. 2

  3. 3

  4. 4

  5. 5

  6. 6

  7. 7

Before you start

  • · manager who can handle skip-level feedback without retaliation
  • · CEO/exec time
  • · explicit norm that this is not insurrection, it''s signal
organizational-listeningexecutive-leadershipseries-bseries-cgrowth-stagelate-stage

Play: Late-night management-by-walking-around (9-9:30pm office walks)

Outcome: Cultural change via recognition, not exhortation.

Context: "Work harder" is rejected by smart people ("you work harder"). Recognition by name in all-hands creates visible upside for late-night work. Within 2-3 months, the pattern becomes self-reinforcing.

I''d come back at 9-9:30 pm and whoever''s here, I''m going to talk about their work at the next all-hands. People slowly changed the culture — those around when Dick''s walking around are getting their stuff prioritized.
Dick Costolo
8-12 weeks to behavioral shift per
  1. 1

  2. 2

  3. 3

  4. 4

  5. 5

  6. 6

Before you start

  • · physical co-located office
  • · CEO/founder bandwidth for 2-3 late nights/week
  • · all-hands meeting cadence already established
organizational-culturefounder-leadershipseedseries-aseries-bseries-c

Play: Write-down-the-feedback-night-before (Bill Campbell)

Outcome: 99 of 100 people will fail to deliver honest feedback live without prep.

Context: Live feedback conversations trigger compensatory empathy reflexes. Pre-written feedback bypasses the reflex by removing live composition.

Write down what you want to say, and then say that and don''t say anything else. 99 out of 100 people, if they go into the meeting the next day, they don''t say that. They''re like ''Listen Brian, it''s really a hard time…'' Just shut the fuck up.
Dick Costolo (citing Bill Campbell)
15-20 min prep night before; 20-30 min meeting per
  1. 1

  2. 2

  3. 3

  4. 4

  5. 5

  6. 6

Before you start

  • · mental rehearsal
  • · commitment to not soften live
  • · willingness to sit with silence
performance-managementexecutive-leadershipseries-aseries-bseries-cgrowth-stagelate-stage

Bias-to-yes — only direct manager or legal can say no

Outcome: Cross-functional veto rights create implicit chains of approval that block work without anyone saying "no" explicitly; restricting veto rights to direct-manager + legal removes the implicit-approval graph and unblocks the work surface.

We're going to have a bias to yes. And bias to yes is: only the person you report to is allowed to tell you you're not allowed to do that. Or legal can tell you if you do that, you'll be fired because it's illegal and a violation of the privacy policy. But no other organization, period, is allowed to tell you you can't do that or you need to go ask these people for permission. Just your direct manager.
Dick Costolo
Standing rule; takes 6-12 months to fully internalize per (proposed)
  1. 1

    State the rule explicitly at an all-hands

    "Only your direct manager (or legal) can tell you no. No other team, no fiefdom, no approval chain. Bias to yes."

  2. 2

    Audit existing approval chains and abolish them

    Find every cross-functional approval that isn't legal-grade. Abolish them.

  3. 3

    Empower experimentation with low-blast-radius defaults

    Engineers can ship to 1% of users without design wireframes. Two-way doors don't need committee.

  4. 4

    Repeat at every all-hands

    Fiefdoms re-form naturally. Re-state monthly until the new norm is internalized.

  5. 5

    Pair with no-punishment-for-mistakes policy

    Bias to yes only works if mistakes don't produce blame.

Stop or pivot when

  • Only direct manager + legal can veto
  • No cross-functional approval chains
  • No punishment for honest mistakes

Scripts

Before you start

  • · CEO willing to publicly defend the rule against fiefdoms re-forming
  • · No-punishment-for-mistakes culture
  • · DRI ownership clear so manager-veto is meaningful
operating-cadenceorganisational-designengineering-cultureseries-bseries-cgrowth-stagepublic-company

Decision Moments

Actual decisions, real outcomes

Specific decisions narrated in the episode with their outcomes and transferable lessons.

Twitter pre-Costolo was a "collective" making decisions by group vote. CEO succession chaotic: Ev pushed out; Costolo informed in driving rain; multiple reversals; Bijan Sabet "Don''t quit!"

Did: Costolo took CEO seat. Killed committee decision-making, banned "so-and-so said" reasoning, removed operating-principles whiteboard + "we''ll make better mistakes" poster. Pushed decisions down the stack.Outcome: Velocity visibly increased. Experimentation rate climbed. Twitter scaled to IPO under Costolo''s 5-year tenure.

Push decisions down the stack. Kill committee culture. Never invoke "so-and-so said" — context, not authority.

Part of an emerging decision pattern across multiple episodes

2011: Twitter is at ~$5B valuation. Kevin Systrom and Mike Krieger built Instagram. 8-9 employees. Costolo recognized Kevin as one of best product thinkers in the Valley.

Did: Offered Instagram $700M stock + 10% of Twitter to acquire. Could have offered another $300M cash via JP Morgan debt. Did not. Facebook acquired Instagram for $1B ($700M stock + $300M cash) a month later.Outcome: Lost Instagram to Facebook. Instagram became the rich-media social moat that boxed Twitter out for the next decade. Vine wasn''t enough.

When you can identify the future of a market, pay anything to acquire it. The cash-borrowing option is the difference between winning and being boxed out.

Public company life at Twitter: every miss-the-revenue-number-by-1% produces employee panic about stock-price-after-hours.

Did: Costolo''s post-hoc framing: "I would have kept it private longer if I could go back and do anything over again."Outcome: Twitter IPO''d 2013; lived through public-company stress for the rest of Costolo''s tenure.

Private companies optimize for long-term over short-term naturally. Going public introduces a stress that doesn''t need to be incurred until necessary.

Tensions surfaced

Contradictions and trade-offs the episode raises — judgment calls a thoughtful operator has to navigate.

Tension

Tension: Founder-mode (Chesky) vs Operator-mode (Costolo)

Founder-mode preserves vision-velocity at cost of scale-decisions; operator-mode preserves scale-decisions at cost of vision-velocity.

Resolution: depends on stage AND founder''s specific strengths. Pre-PMF + design-led product (Chesky''s Airbnb) = founder-mode. Post-PMF + 1000+ headcount + execution-heavy ops (Costolo''s Twitter) = operator-mode.

Push decisions down the stack. Don''t come to me and ask what you should do unless you have no idea. Figure it out. Push as many decisions down the stack as you can.Dick Costolo

Durability: Durable. The "operating mode must match decision cost structure" pattern is structural.

Productive tension named in episode title.

Tension

Tension: Process as solution vs DRIs and operating control

Both real — but process-creep dominates DRI-creation in mature orgs. Default trajectory is over-process.

Process = permanent additive cost. DRI = recurring judgment cost. Processes are easier to install and harder to remove. Default org momentum builds processes.

Don''t solve problems with processes. Google''s launch checklist got to 17 pages. Pretty soon you''re managing to processes not outcomes. Solve with DRIs, operating control, not processes.Dick Costolo

Durability: Durable. The "processes accumulate, DRIs erode" pattern is structural.

Productive tension with explicit resolution.

Corpus connection

Where this episode fits for retrieval

What kinds of decisions this briefing is best pulled into.

Primary decisions

  • operating-cadence
  • executive-hire
  • organisational-design

Temporal flag

timeless