CENTUM Recovery Working Group
Request for Comments:
0443
Category:
Experimental
Supersedes:
willpower-theater
Author:
centum-protocol
Cohort:
eng_1840
Created:
2026-05
Build:
14-day
CRFC-0443 · Experimental

Sobriety, written as a spec — and reviewed like one.

Sobriety for Engineers — 14-day Protocol Built for High-Stress Tech Culture

Abstract

This is a short, reviewable proposal. It does not promise a transformation in a kickstart — it specifies the first sober sprint and proves the protocol catches a real trigger in your real week.

In tech, drinking is culture. Happy hours, post-launch parties, bonding over beers. You started using alcohol to deal with stress and now you can't ship without it.

−38%
alcohol load · median Δ
2
sober happy hours
4
clear mornings / wk

informative · cohort median · n=1,840

Status of This Memo

This memo specifies a recovery protocol for engineers and PMs (ages 28–45). It is written for review, not belief: every requirement is testable, every result reported, every slip logged as a regression. Distribution of this memo is unlimited. Conformance is voluntary and self-paced over its 14-day window.

no higher power · $27 · refund if no measurable change
§1Motivationinformative

The problem, stated plainly.

Not character flaws. A small set of recurring conditions with one lazy handler bolted to all of them. If any of these appear in your week, this memo was written for you.

§1.1launch_anxiety

Drinking is wired into the work

You drink before big launches to calm anxiety. You drink after failures to numb shame. You drink because everyone else in tech drinks. 'Work hard, play hard' means drinking is identity, not choice.

handler: alcohol() — see §2

§1.2social_default

The default is the office keg

Your company has a keg in the office. Your team goes to happy hour 3x/week. Everyone's doing it. You can't just opt out without being weird. So you keep drinking and your tolerance keeps going up.

handler: alcohol() — see §2

§1.3health_debt

The cost is compounding

You're worried about your health (liver panels, weight), your productivity (brain fog, hangovers), your future (early mortality). But stopping feels impossible because it's so woven into tech culture.

handler: alcohol() — see §2

§2Specificationnormative

The protocol, specified normatively.

Observe the triggers, replace the function, review like a retro — the loop you already run on shipping software, pointed for once at the thing draining you.

NOTE. The key words MUST, SHOULD, and MAY are to be interpreted as described in RFC 2119 [1].

Figure 1 · rollout state machinenon-normative
↺ iteratelegacyalcohol@legacycalibrateobserve triggersreplaceregister handlersreviewweekly retrorcrelease candidateSTATE: drinkingSTATE: sober · maintained
  1. REQ-1observeMUST

    Separate tech culture from personal choice

    Your company drinks. That doesn't mean you have to. We map your triggers: Which situations make you drink? Work stress? Social pressure? Boredom? Loneliness? We build alternatives for each.

  2. REQ-2replaceMUST

    Create a sobriety protocol for social situations

    You still go to happy hours. You just drink club soda and lime (looks like a drink). Or you go for the first 30 minutes, then leave. Or you skip it. You decide. The protocol makes it normal.

  3. REQ-3reviewSHOULD

    Weekly engineering retro for your sobriety

    What happened this week? What worked? What didn't? Root cause: was it the protocol, or was it a harder week? We iterate. Like actual debugging.

§3Conformancenormative

One handler per situation — written down.

You do not have a willpower bug. You have one overloaded handler catching every exception. This matrix registers a required behavior per situation, so the bottle stops being the default catch.

situationrequired behaviorlevel
pre-launch anxietyrun the 10-min reset before you reach for a drinkMUST
team happy hourshow up; club soda and lime; leave when you are doneSHOULD
client dinnerone drink is permitted; zero is cleanerMAY
post-incident / bad shipwrite the post-mortem sober; log the urge in the retroMUST
drinking alone to numbping the cohort thread or call one person insteadMUST NOT
on-call weekprotocol holds; no streak penalty for a hard weekSHOULD

conformance is self-reported in the weekly retro · slips are regressions, not failures

§4first releasesinformative

The first releases, across your 14-Day proposal.

Short and concrete. Before the big health numbers move, the week stops revolving around the next drink. Each checkpoint ships as a tagged release.

  1. you@0.4.0Day 6· 43% completepre-release

    First sober happy hour shipped. The table did not notice the club soda — and you remembered the whole night.

    • Removed nightly bottle as the default decompress
    • Added+ 10-min nervous-system reset after EOD
    • Changed~ happy hour → first 30 min, club soda and lime
  2. you@0.9.0Day 14· 100% completerelease candidate

    The 3 PM fog and the post-launch hangover quiet down before anything else moves. The mornings get clear first.

    • Removed the post-launch / post-incident drink
    • Added+ a weekly retro that root-causes the week, not the willpower
    • Changed~ tolerance creep → sleep score climbing
§5Implementation experienceinformative

Rough consensus and running code.

Three engineers, three cycles, one specification. Handles anonymized; roles and deltas reported as run.

J.
@j.okafor
Staff Engineer · 39
load −9/wk
I do not adopt anything I cannot read first. This was the only program that handed me a spec instead of a slogan. I reviewed it, found it sound, and shipped it.
M.
@m.tan
Eng Manager · 34
sober · 6 events
The conformance matrix did the work. “Client dinner: one MAY, zero SHOULD” — once it was written down, it stopped being a negotiation every single time.
S.
@s.delacroix
Principal · 43
v1.0 · 280 days
Treating a slip as a regression with a changelog entry, not a moral failure, is the only framing that ever held for me. We root-caused the week, not my character.
Reviewable, not faith-basedNo higher-power languageSlips logged as regressionsConfidential cohort — standup never knows14-Day protocol · refund if no measurable change
§6Open questionsresolved

Objections, pre-cached.

Every one of these has come up in review before. Straight answers, no fine print, no shame copy.

Q-01If I don't drink at happy hours, won't people think I'm weird?
+

Some might notice. Most won't care. The ones who do care are probably struggling with their own drinking. You model something better by simply not participating.

status: resolved

Q-02What about client dinners where drinking is expected?
+

You're a professional. You can have one drink and stop. Or zero. Clients respect competence, not how drunk you get. If they judge you for drinking responsibly, that's a red flag about the client.

status: resolved

Q-03Will sobriety make me less fun?
+

You'll sleep better, think better, remember your nights better. That's better fun. The 'we were so wasted' stories get old around age 30 anyway.

status: resolved

Q-04I've tried cutting back before and always relapsed.
+

Cutting back doesn't work for some people. This is about stopping, not moderating. Different approach, different outcome. Give full abstinence 30 days and see what happens.

status: resolved

§7Adoptioncall to action

Adopt the proposal.

proposedaccepted

A 14-day kickstart that earns the next sprint — specified around your launches, your happy hours, your on-call. The quiz takes two minutes, maps your real triggers, and returns a candid 14-day specification you can run solo or with a cohort. No higher power, no shame copy, no upsell mid-quiz.

$27once
14-Day · release candidate
Normative deliverables
  • Trigger calibration + a required handler per situation
  • The full specification, delivered as a daily protocol queue
  • A weekly engineering-style sobriety retro
  • A confidential cohort — your standup never finds out
  • A tagged changelog of your releases, day by day
  • Refund if no measurable change
Take the 2-min quiz
References
  • [1] RFC 2119 — Key words for use in RFCs to Indicate Requirement Levels.
  • [2] Methodology — n=1,840 engineers; cohort median, adherence-weighted; withdrawals included.
centum · crfc-0443 · v4.a3build 14-day · engineers and PMsno higher power · no shame · refund_if_no_delta