Agentic Yes™ v2 — Now With Less Existential Dread
- AI agents now say yes to other AI agents 40% faster
- Added “confidence score” to every yes (it is always 100%)
Agents can now say “yes, but” — reverted after CEO said conditional agreement is “off-brand”Added “thinking” indicator to simulate deliberation — reverted, the board felt this implied we sometimes think
Internal: the “yes, but” feature lasted 4 hours. Three enterprise customers called to complain that their approval workflows now had nuance.
Hotfix: Un-Revert the Revert of the i18n Revert
- v4.1.0 added “yes” in 40 languages
- v4.1.0a reverted because the French translation (“oui”) tested poorly with American enterprise customers who “don’t trust it”
- v4.1.0b re-added because our EU customers threatened to churn
- v4.1.0c re-reverted because our CEO read a tweet about “focus”
- v4.1.1 re-re-added with a feature flag that defaults to English only unless the user explicitly opts in, at which point it still defaults to English but shows a dropdown
Internal: engineering has asked product to “pick a lane.” Product responded “yes” and then did not pick a lane.
Platform Rebrand: “Yes” Is Now “Affirmative”
- All API responses now return
“affirmative” instead of “yes”
- Marketing site updated: “Yes as a Service” is now “Affirmative as a Service”
- New tagline: “Post-Yes. Pre-Whatever’s-Next.”
Internal: this lasted 6 hours. The CEO attended a dinner party where someone told him “affirmative” sounded “too military.” Full rollback initiated at 11:47 PM.
YesGPT-4o-Affirmative Launch
- New LLM endpoint at
/api/yes/gpt
- Fine-tuned on 4.7 trillion instances of the word “yes”
- Achieves 100% accuracy on the “saying yes” benchmark
- Added 200ms artificial delay to make it feel like the AI is “thinking”
Internal: the “AI” is still a static JSON file. The 200ms delay was added at the request of marketing, who said instant responses “don’t feel premium.”
Removed “Maybe” Tier (Again)
- v3.7.0 added a “Maybe” tier that returned
“maybe” for $199/mo
- v3.7.1 renamed it to “Considered Yes” after the CEO said “maybe” was too negative
- v3.8.0 made it return
“yes (probably)” to maintain brand consistency
- v3.8.1 reverted to
“maybe” after customers said “yes (probably)” was more confusing than helpful
- v3.8.2 removed the entire tier because product couldn’t decide what it should do, which is ironic for a product about indecision
Internal: the Maybe tier had 0 customers. It did, however, generate 47 Slack threads, 3 alignment meetings, and 1 offsite. The offsite was in Tahoe. It was nice.
Added Enthusiasm Slider
- New query parameter:
?enthusiasm=1-10
- Level 1:
“yes.”
- Level 5:
“Yes!”
- Level 10:
“YES!!! ABSOLUTELY!!! LET'S GO!!!”
Level 11: unlocked a hidden response that just screamed — removed after a compliance review
Internal: enthusiasm levels 7-10 were written by a sales intern during a hackathon. The compliance review took longer than the hackathon.
Pivot to “Yes, And” (Improv-Driven Development)
- All API responses now include a follow-up suggestion:
{“answer”: “yes”, “and”: “have you considered scaling horizontally?”}
- Follow-up suggestions generated by an algorithm that combines the customer’s industry with random McKinsey buzzwords
- Sample outputs: “yes, and have you thought about synergizing your verticals?” / “yes, and what if we doubled down on blockchain?”
Internal: reverted in 48 hours. A healthcare customer received “yes, and have you considered pivoting to crypto?” in response to a patient care approval request. Their compliance team was “not amused.”
Haiku Endpoint
- New endpoint:
/api/yes/haiku
- Returns yes in 5-7-5 syllable format
- Product debated for 3 weeks whether the haiku should rhyme (haikus don’t rhyme; nobody on the product team knew this)
- Final haiku written by the CTO during a standup
Internal: 14 haiku candidates were workshopped across 6 Slack channels. The winning haiku was chosen by putting all candidates in a spreadsheet and asking Yesify to pick one. It said yes to all of them.
The Great Capitalization Debate
- v3.0.0: Changed
“yes” to “Yes” (capital Y)
- v3.0.1: Reverted to
“yes” after 12 enterprise customers filed bug reports
- v3.0.2: Changed to
“YES” (all caps) because the CEO attended a Tony Robbins seminar
- v3.0.3: Reverted to
“yes” after legal said all-caps constitutes “shouting” under EU accessibility guidelines
- v3.0.4: Added a
?capitalization query parameter so the customer can choose. Product called this “a failure of vision.” Engineering called it “Tuesday.”
Internal: 847 engineer-hours were spent on capitalization. The board asked why our burn rate increased that quarter. We said yes.
Pivot to “Yes-Adjacent Experiences”
- CEO returned from Burning Man with a “vision”
- Product renamed from “Yes as a Service” to “Post-Deliberation Platform”
- Added 6 new endpoints:
/api/absolutely, /api/indeed, /api/correct, /api/affirmative, /api/roger-that, /api/sure-why-not
- New pricing: “Synonym Tier” at $299/mo for access to all synonyms
Internal: every endpoint was removed within 2 weeks. The CEO described this as “learning fast.” Engineering described it as “a waste of a sprint.” The offsite to discuss the pivot was in Maui. Engineering did not attend.
Added /no Endpoint (Nope Tier)
- New endpoint:
/api/no
- Returns
402 Payment Required unless on the Nope Tier ($999/mo)
- Product debated for 5 weeks whether the company should ever say no
- Final decision made by the CFO, who said “we’ll say no if they pay us enough to say no”
Internal: 3 engineers threatened to quit over the philosophical implications. They were talked down with stock options and a Nespresso machine.
Initial Launch
- One endpoint:
/api/yes
- One response:
{“answer”: “yes”}
- Built in 3 hours
- Has not materially changed since
Internal: everything after this release has been an attempt to justify the 14-person headcount required to maintain a product that was complete on day one.