{ "flags": [ { "id": "writing-needs-reader-context", "meeting_id": "SPM16-02", "kind": "worth_spreading", "title": "Person Q’s model drafts name four agents without explaining who they are", "face": "Testing model-written email and social pitches, Person Q found drafts that assumed you already knew her context. One email named Claudie, Andy, Viktor and Fin without explaining whether they were people or agents; benchmark percentages arrived without baselines. Her diagnosis: “the pitches are all very navel-gazey - buried in my own context, not extracting what a reader can learn from or apply”.", "context": "In September 16 Slack writing tests, Person Q critiques social pitches and an email draft for failing to supply reader context. These are her reported judgments; the attached drafts were not independently inspected.", "moment": [ { "line": 23, "speaker": "Person Q", "quote": "the pitches are all very navel-gazey - buried in my own context, not extracting what a reader can learn from or apply" }, { "line": 76, "speaker": "Person Q", "quote": "Claudie, Andy, Viktor, Fin — four names with no hint they're agents, systems, or people." }, { "line": 77, "speaker": "Person Q", "quote": "The benchmark numbers — 44, 85, 80 percent, and \"62\" all land without a baseline or a \"up from what.\"" } ], "why": "You get concrete examples for judging whether a draft supplies the context its intended reader needs: identify unfamiliar names and give numbers a baseline.", "people": [ "Person Q" ], "options": [], "strategy_quote": null, "reaction_inferred": null, "confidence": 0.92 }, { "id": "skills-adoption-metric-tension", "meeting_id": "NPM16-04", "kind": "room_texture", "title": "Person B and Person D question whether more skill use would mean a better Agent", "face": "In the Agent roadmap review, likely Person B challenged whether better skills would get more people to accept automations: “I don't want us to fake like being focused.” Likely Person D agreed that more skill use without better satisfaction, revenue or memberships would be a failure. Both still supported skills as infrastructure; the disputed link was whether the proposed work would improve adoption.", "context": "The September 16 product review was examining the roadmap for automation activation when the conversation turned to whether skill improvements would move that metric. Attribution is tentative: the supplied two-person roster, roadmap presentation and sustained question-and-answer turns support Person D presenting and Person B challenging; the transcript has no speaker labels.", "moment": [ { "line": 103, "speaker": "likely Person B", "quote": "I don't want us to fake like being focused." }, { "line": 106, "speaker": "likely Person D", "quote": "If we ship a bunch of updates to Skills and then Skill usage goes up because we like, you know, orient around that. But. like overall satisfaction, revenue. Memberships? Don't change. then that's a failure." }, { "line": 107, "speaker": "likely Person D", "quote": "What I do believe though is that skills are a useful primitive, to use your word, to as a delivery mechanism for value that the agent could provide." } ], "why": "You get the substantive tension behind the roadmap: both support skills, while questioning whether skill usage would demonstrate customer value.", "people": [ "Person B", "Person D" ], "options": [], "strategy_quote": null, "reaction_inferred": null, "confidence": 0.84 }, { "id": "diagnostic-count-ban-erases-timings", "meeting_id": "GPM16-03", "kind": "worth_spreading", "title": "Engineer A’s prompt ban on diagnostic counts also erased useful timings", "face": "Engineer A’s PR reports that telling the Agent to omit diagnostic counts also made it drop useful timings—250ms, 120ms and 8s—from unrelated replies. Adding “never overrides the numbers rule above” scored worst: zero of three runs passed. The lesson you can test: a narrow prohibition may affect other answers. The revised prompt removes the count ban; a clean final evaluation is still pending.", "context": "The September 16 [retry-and-clarity PR #1755](https://sources.example.invalid/reference-050) records regressions while trying to translate an external copy-edit service’s diagnostics into plain language. The complete discussion and afternoon head commit retain the removal of the count ban; the PR is open, and its Claude-assisted writeup reports small-sample results rather than an independently reproduced general law.", "moment": [ { "line": 70, "speaker": "engineer-a", "quote": "the planted timings (250ms, 120ms, 8s) dissolved into qualitative phrases **in replies where nothing had failed and no tool had been called**." }, { "line": 75, "speaker": "engineer-a", "quote": "Priming does not read disclaimers; a reassurance names the class as loudly as a ban." } ], "why": "You get a concrete failure to test when refining writing instructions: a ban on one kind of detail can remove useful detail elsewhere, even with an explicit exception.", "people": [ "Engineer A" ], "options": [], "strategy_quote": null, "reaction_inferred": null, "confidence": 0.85 }, { "id": "golf-pelican-and-sourdough-sword", "meeting_id": "SPM16-01", "kind": "room_texture", "title": "Person N’s one-prompt games added a ball-stealing pelican and a sourdough sword", "face": "Person N reported his most thorough one-prompt golf game yet: a flying course preview, a camera tracking the ball, and a pelican that steals it. Another game supplied a sourdough sword and $4,200 rent. His reaction gives you the texture of the test: “is this really opus? this is so good for one prompt.” Person P liked it too; Person N said Astra hadn’t come close in his personal testing.", "context": "In September 16 Slack exchanges, Person N shares game videos and explains which details delighted him; Person P responds to the second example. The videos and prompts were not inspected, and the model comparison is Person N’s personal report.", "moment": [ { "line": 6, "speaker": "person-n", "quote": "i do this thing where i one-shot golf games and this is was the most thorough ive ever seen. it seems like it did not research the flying preview and post-shot camera change, and that was a really impressive detail to me that it figured out on its own" }, { "line": 13, "speaker": "person-n", "quote": "• a pelican that occasionally steals my ball :skull:" }, { "line": 70, "speaker": "person-n", "quote": "is this really opus? this is so good for one prompt. and love the creativity w/ the sourdough starter buddy, the $4200 rent, the sourdough bread sword." } ], "why": "You can see exactly what excited Person N: cinematic camera choices and playful details in his game tests, without treating his enthusiasm as a measured model ranking.", "people": [ "Person N", "Person P" ], "options": [], "strategy_quote": null, "reaction_inferred": null, "confidence": 0.8 } ], "reading_notes": [ { "meeting_id": "NPM16-01", "substance": "The launch team works through skills examples, compounding copy, permissions illustrations and confusing pricing.", "selection_reason": "Omitted: the concrete revisions have owners, and the remaining pricing work is already routed rather than an unresolved CEO tiebreaker." }, { "meeting_id": "NPM16-02", "substance": "Ghostwriter Person K accepts a call-prep automation after discussing Claude context, private client conversations and administrative work.", "selection_reason": "Omitted: useful onboarding feedback, but no captured AI-ambassador pitch and reaction or observed automation result supports a stronger flag." }, { "meeting_id": "NPM16-03", "substance": "Person L and Person E plan nontechnical Office Hours, discuss reusable video skills and compare their ongoing tool-building experiments.", "selection_reason": "Omitted in favor of the selected moments; the Remotion reuse versus discoverability exchange is concrete, while the Agent pitch is internal and the attention-workflow account repeats familiar material." }, { "meeting_id": "NPM16-04", "substance": "Person B and Person D examine automation activation, skills, growth experiments, billing friction and engineering specifications.", "selection_reason": "Selected the tension over whether skill improvements would move customer outcomes; omitted the previously surfaced activation rate and routine roadmap status." }, { "meeting_id": "NPM16-05", "substance": "UC Davis IT director Person M tries onboarding in a roughly 30,000-person Slack workspace and asks for private-channel testing and clear data boundaries.", "selection_reason": "Omitted: this is cautious onboarding feedback with no demonstrated useful outcome or qualifying external reaction to the ambassador message." }, { "meeting_id": "SPM16-01", "substance": "Person N shares one-prompt games whose camera behavior and comic details delight him and draw positive reactions.", "selection_reason": "Selected the concrete creative details and his excitement, preserving the personal-test limitation." }, { "meeting_id": "SPM16-02", "substance": "Person Q tests writing and social pitches, identifying missing reader context, weak coherence and instruction-following problems.", "selection_reason": "Selected the concrete reader-context critique; did not turn separate model slowness reports into a shared cause." }, { "meeting_id": "SPM16-03", "substance": "The team audits retention and vendor disclosures, and Person B ultimately specifies 90 days while the privacy policy is reworked.", "selection_reason": "Omitted: the apparent 14-versus-90-day dispute is resolved later in the thread, and remediation has named owners without a decision waiting on you." }, { "meeting_id": "SPM16-04", "substance": "A brief API outage prompts an install-load hypothesis before Person J identifies a restart caused by an environment change.", "selection_reason": "Omitted: routine incident follow-up is owned, and the initial large-install explanation is qualified by later evidence." }, { "meeting_id": "SPM16-05", "substance": "Person J announces Slack agent mode and clarifies that the native Stop button appears in the agent DM panel while channels accept typed stop.", "selection_reason": "Omitted: you reacted to the announcement, and the later interface clarification resolves the immediate confusion." }, { "meeting_id": "SPM16-06", "substance": "Person S shares your benchmark, explains ownership transfer, and Person T takes the rendering issue after your hands-on feedback.", "selection_reason": "Omitted: you participated throughout, and the next steps and repair already have routes." }, { "meeting_id": "SPM16-07", "substance": "Person T describes evals as replaying past work, colleagues extend your reference-check analogy, and you request audience-relative checks.", "selection_reason": "Omitted: the central new request is yours, and the replies add little beyond the discussion you joined." }, { "meeting_id": "SPM16-08", "substance": "The tax integration discussion covers Stripe checkout, central imports and a correction that app histories are still awaiting export.", "selection_reason": "Omitted: the correction matters operationally but remains in an owned implementation thread rather than one of the five reading moments." }, { "meeting_id": "SPM16-09", "substance": "Person L coordinates the September 25 Agent preview and October camp, with Person I and Person B offering participation.", "selection_reason": "Omitted: the uncertainty is substantially resolved through ordinary event coordination." }, { "meeting_id": "SPM16-10", "substance": "Person N and Person H compare flat and dimensional UI treatments, then converge on sharp corners for cards and buttons.", "selection_reason": "Omitted: the later agreement resolves the design choice, and the brief exchange adds limited texture beyond implementation detail." }, { "meeting_id": "SPM16-11", "substance": "Person Q says Vibe Check load and changes to the Compound Engineering story call for more time; Person G asks how much.", "selection_reason": "Omitted: a scheduling question remains, but no settled delay or substantial tiebreaker is established." }, { "meeting_id": "SPM16-12", "substance": "Person F finds it confusing that pausing a goal can leave task agents working, in a discussion you join about Quasar goals.", "selection_reason": "Omitted: the exchange is familiar to you and supplies no distinct additional perspective beyond the reported UX mismatch." }, { "meeting_id": "SPM16-13", "substance": "Person I says sponsor opt-in is live for the afterparty and awaits sponsor names for the next setup.", "selection_reason": "Omitted: the operational dependency is between named colleagues and does not support a reading flag." }, { "meeting_id": "SPM16-14", "substance": "Person T reports Person AG’s interest in trying the Agent and asks how to enroll him.", "selection_reason": "Omitted: secondhand interest contains neither the pitch nor a directly captured external reaction." }, { "meeting_id": "SPM16-15", "substance": "Person D builds on Person I’s department-onboarding proposal with default job-title inference and optional installer selection.", "selection_reason": "Omitted: this remains an uncommitted product idea rather than a distinct qualifying exchange in the final small selection." }, { "meeting_id": "GPM16-01", "substance": "Person S closes the codebase benchmark import and merges an MCP path for a new account’s agent to create its benchmark.", "selection_reason": "Omitted: a concrete architecture choice is already settled, and your related benchmark thread makes a simple progress recap less valuable." }, { "meeting_id": "GPM16-02", "substance": "Person D’s open reply-formatting PR documents worked examples, eval/live-context parity and an exception for important unasked facts.", "selection_reason": "Omitted: the afternoon delta is approval of September 14–15 work; the newer prompting regression provides the stronger distinct reading moment." }, { "meeting_id": "GPM16-03", "substance": "Engineer A’s open PR adds bounded retry for degraded success and records how diagnostic-count prohibitions affected unrelated numerical answers.", "selection_reason": "Selected the reported prompting regression as a reusable finding, retaining small samples and pending final evaluation rather than claiming a shipped fix." }, { "meeting_id": "GPM16-04", "substance": "Person P’s merged external-agent profile change separates self-reported capabilities from observed behavior.", "selection_reason": "Omitted from the four-card selection: the distinction is useful, but the supplied material is primarily a schema and profile-alignment change." }, { "meeting_id": "GPM16-05", "substance": "Person P’s merged handoff lets Baby wait for another Slack bot and resume the same turn, including a fix for replies discarded after roster changes.", "selection_reason": "Omitted: this is implementation progress with a resolved review finding and no observed real-model outcome or committed next experiment." } ], "notes": "All 25 supplied primary source units were read in full, including later qualifications and relevant patches. Notion transcripts are unlabeled and sometimes garbled; roadmap attribution remains likely, supported by the two-person roster and local turn continuity. Slack authors and GitHub accounts are identified; account attribution does not establish human prose authorship. Quoted spans retain original characters and L-locators. Linked attachments were not inspected. No attendance or novelty is assumed. The strategy is a dated reference; its promised message addendum is absent from the supplied reference file, and no message_tested or off_strategy claim is made. Source links and provenance remain in the packet; all output is draft-only." }