{ "flags": [ { "id": "sep16-q-social-pitches", "meeting_id": "SPM16-02", "kind": "worth_spreading", "title": "Person Q’s writing lens couldn’t rescue Marzipan’s social pitches", "face": "Person Q asked Marzipan for social posts she would actually use, then tried her homebrewed Compound Writing lens based on your judgment. Even after several rounds, her verdict stayed sharp: “the pitches are all very navel-gazey - buried in my own context, not extracting what a reader can learn from or apply”. She says the lens usually finds good angles; here, it didn’t rescue the argument.", "context": "In September 16 writing-test Slack discussions, Person Q tested whether Marzipan could pitch usable social ideas and described repeated attempts to improve its angles. Source: https://sources.example.invalid/reference-032, L6–30; her reports and attached screenshots are not an independently reproduced evaluation.", "moment": [ { "line": 6, "speaker": "Person Q", "quote": "Running a new experiment: Can the model pitch me ideas for my social that I would actually use? Running on Marzipan now" }, { "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": 30, "speaker": "Person Q", "quote": "I even gave it my homebrewed Reader A lens skill from compound writing, which i've found to be very good at finding angles" } ], "why": "You get a precise writing failure to think with: a model can stay immersed in the author’s context without finding anything the audience can use.", "people": [ "Person Q" ], "options": [], "strategy_quote": null, "reaction_inferred": null, "confidence": 0.91 }, { "id": "sep16-diagnostic-counts-timings", "meeting_id": "GPM16-03", "kind": "worth_spreading", "title": "A ban on diagnostic counts made the Agent drop useful timings elsewhere", "face": "Engineer A’s PR reports a trap for your prompt-writing: telling the Agent to omit a provider’s diagnostic counts also made it drop useful timings in unrelated replies. An explicit exception scored worst, at 0/3: “Priming does not read disclaimers; a reassurance names the class as loudly as a ban.” These are small reported tests; the replacement wording still awaits its final evaluation.", "context": "In the open September 16 retry PR, attempts to suppress confusing copy-edit diagnostics regressed an unrelated timing-preservation case: the PR reports a 3/3 baseline and variants at 1/3, 0/3 and 1/3. Primary evidence: https://sources.example.invalid/reference-050, L60–99, and afternoon head https://sources.example.invalid/reference-045; attribution is to the engineer-a account’s AI-assisted PR, not independently verified human authorship or a universal causal result.", "moment": [ { "line": 70, "speaker": "engineer-a (PR description)", "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 (PR description)", "quote": "the clause literally said it *never overrides the numbers rule above*. This scored **worst (0/3)**." }, { "line": 75, "speaker": "engineer-a (PR description)", "quote": "Priming does not read disclaimers; a reassurance names the class as loudly as a ban." } ], "why": "You get a concrete counterexample to assuming a narrow instruction only changes the situation it names, with enough reported test detail to keep the lesson provisional.", "people": [ "Engineer A" ], "options": [], "strategy_quote": null, "reaction_inferred": null, "confidence": 0.9 }, { "id": "sep16-getting-replies", "meeting_id": "NPM16-03", "kind": "room_texture", "title": "Person L asks how to get replies; Person E says he keeps asking", "face": "While planning Agent office hours, likely Person L asked how to get colleagues to answer: “I feel like I'm constantly tagging and DMing people and just often never hear back.” Likely Person E said he asks Person B and Person D every day until they respond, or catches them on calls. You get the practical texture here: repeated nudges are how these two describe getting answers.", "context": "In the September 16 Person L × Person E meeting, discussion of inviting a technical colleague to office hours turned into an exchange about getting responses (https://sources.example.invalid/reference-005, L89–111). The transcript is unlabelled: the two-person roster, Person L’s hosting remarks, Person E’s self-reference at L84 and launch work at L109, and their alternating exchange support these tentative local attributions; their accounts do not independently establish anyone’s overall responsiveness.", "moment": [ { "line": 97, "speaker": "likely Person L", "quote": "How do you get answers? Yeah, I feel like I'm constantly tagging and DMing people and just often never hear back." }, { "line": 104, "speaker": "likely Person E", "quote": "Like, Person B and Person D have been tough sometimes for me in terms of the every agent. Like, when I have questions about things, like... I just ask, I'll be asking them like every day on Slack until they answer me." }, { "line": 107, "speaker": "likely Person E", "quote": "Or if they're on a call, I'll ask them on the call." } ], "why": "You hear the extra effort colleagues describe spending to get answers, without turning their frustration into a conclusion about anyone’s performance.", "people": [ "Person L", "Person E", "Person B", "Person D" ], "options": [], "strategy_quote": null, "reaction_inferred": null, "confidence": 0.81 }, { "id": "sep16-agent-claims-observations", "meeting_id": "GPM16-04", "kind": "worth_spreading", "title": "Person P’s agent profiles separate claimed abilities from observed behavior", "face": "For agents you delegate to, Person P’s merged Baby Agent change preserves two records: what a bot says it can do, and what happened when it got work. The rule is explicit: “An agent's self-description is a claim, not evidence.” A slow, wrong or blocked handoff goes into Observed behavior without rewriting the bot’s answers; a profile with no observations says so.", "context": "Person P’s September 16 Baby Agent PR #204 aligns external-agent profiles with his interview skill and keeps observed behavior separate from claimed capabilities (https://sources.example.invalid/reference-038, L13–30, L159–165 and the model patch at L192–209). The PR is merged, while deployment and production imports remain listed next steps; the card surfaces the design rule rather than live availability or demonstrated routing improvement.", "moment": [ { "line": 161, "speaker": "person-p (PR documentation)", "quote": "An agent's self-description is a claim, not evidence." } ], "why": "You get a concrete way to keep a persuasive agent introduction from becoming assumed evidence of competence in later delegation.", "people": [ "Person P" ], "options": [], "strategy_quote": null, "reaction_inferred": null, "confidence": 0.8 } ], "reading_notes": [ { "meeting_id": "NPM16-01", "substance": "The launch team revised the landing page’s skill examples, compounding explanation, permissions demo and membership presentation.", "selection_reason": "Omitted because the concrete disagreements mostly acquired a direction or follow-up, and the remaining launch work adds less than the selected moments." }, { "meeting_id": "NPM16-02", "substance": "Ghostwriter Person K installed the Agent, accepted a call-prep automation and described interest in client administration after earlier automations disappointed him.", "selection_reason": "Omitted because the call does not test the company-wide show-not-tell message, and the accepted automation has no observed result yet." }, { "meeting_id": "NPM16-03", "substance": "Person L and Person E planned nontechnical Agent office hours, discussed getting colleagues’ replies, and compared video-making workflows.", "selection_reason": "Selected their candid exchange about repeated nudging as room texture; omitted the familiar pitch and your attention workflow." }, { "meeting_id": "NPM16-04", "substance": "Person B and Person D challenged the causal link between skills work and automation activation, then discussed experiments, billing friction and engineering specifications.", "selection_reason": "Omitted because the discussion adds nuance but leaves no clear decision for you, while the activation rate and skills-consumer theme are already represented in prior cards." }, { "meeting_id": "NPM16-05", "substance": "UC Davis IT director Person M installed the Agent in a reportedly 30,000-person Slack workspace and asked about private channels, data access and permissions.", "selection_reason": "Omitted because this is an exploratory onboarding, not demonstrated adoption at that scale or a test of the tracked message." }, { "meeting_id": "SPM16-01", "substance": "Person N shared one-prompt games with inventive camera behavior and comic details, drawing enthusiastic reactions from Person O and Person P.", "selection_reason": "Omitted from the small final selection; the delight is concrete, but the videos were not independently inspected and the selected moments offer more substance." }, { "meeting_id": "SPM16-02", "substance": "Person Q reported social pitches trapped in author context and writing samples missing connective logic, explanations and numerical baselines.", "selection_reason": "Selected the Marzipan social-pitch test and failed lens intervention as a crisp account of what reader-oriented writing requires." }, { "meeting_id": "SPM16-03", "substance": "The retention audit expanded to additional vendors and deletion behavior, and Person B settled the discussion on 90 days while the privacy policy is reworked.", "selection_reason": "Omitted because the apparent 14-versus-90-day split was resolved and follow-up ownership is present; it is not an open tiebreaker or an unowned security escalation." }, { "meeting_id": "SPM16-04", "substance": "After a brief API outage, Person D suspected a large installation, but Person J attributed the restart to an environment-variable change and investigation continued.", "selection_reason": "Omitted as a recovered operational incident with active owners; the later correction prevents presenting the large installation as the established cause." }, { "meeting_id": "SPM16-05", "substance": "Slack Agent mode became available, and confusion about stopping work was resolved into a DM-panel button versus typing stop in channels.", "selection_reason": "Omitted because you already reacted to the release and the subsequent clarification is routine product guidance." }, { "meeting_id": "SPM16-06", "substance": "Person S created your benchmark, explained ownership through Members, and Person T took the rendering issue while directing check edits through MCP.", "selection_reason": "Omitted because you directly participated and the outstanding work already has a route." }, { "meeting_id": "SPM16-07", "substance": "Person T described evals as replaying past work, Person U extended the reference-check analogy, and you requested audience-relative checks.", "selection_reason": "Omitted because the strongest current direction is your own request and the surrounding analogies add little beyond that familiar exchange." }, { "meeting_id": "SPM16-08", "substance": "The tax-integration thread clarified central imports and account cleanup, then corrected the claim that app history was already in Kintsugi.", "selection_reason": "Omitted because the correction preserves the existing no-local-backfill route and does not create a decision for you." }, { "meeting_id": "SPM16-09", "substance": "Person L and Person E kept the All Access preview, while Person I requested product-maker participation and Person B offered to join next Friday.", "selection_reason": "Omitted because the participation question is being resolved and the thread is primarily event coordination." }, { "meeting_id": "SPM16-10", "substance": "Person N and Person H discussed depth and corner treatment, asked for a place to try designs in context, and settled on sharp cards and buttons.", "selection_reason": "Omitted because the visual disagreement resolves in the final reply and the staging preference repeats an existing theme." }, { "meeting_id": "SPM16-11", "substance": "Person Q said the Compound Engineering story needs more time because of Vibe Check load and changes since its initial draft.", "selection_reason": "Omitted because choosing a one- or two-week delay is routine scheduling rather than a meaningful tiebreaker." }, { "meeting_id": "SPM16-12", "substance": "Person F found paused goals confusing because task agents kept working, in a discussion you joined about goals on simple tasks.", "selection_reason": "Omitted because you already responded to the precise observation and a recap supplies little additional value." }, { "meeting_id": "SPM16-13", "substance": "Person I told Person U that the after-party opt-in is live but the broader sponsor-sharing launch is waiting on sponsor names.", "selection_reason": "Omitted because the missing input and person-to-person route are already explicit without your intervention." }, { "meeting_id": "SPM16-14", "substance": "Person T reported that Person AG at a16z wants to test the Agent and asked how beta enrollment works.", "selection_reason": "Omitted because interest and an enrollment question do not yet establish a substantive experiment or an external reaction to the tracked message." }, { "meeting_id": "SPM16-15", "substance": "Person I proposed department-owner interviews during onboarding, and Person D suggested automatic discovery with an optional installer override.", "selection_reason": "Omitted because this remains an exploratory product idea without a committed test or a sharp unresolved choice." }, { "meeting_id": "GPM16-01", "substance": "Person S closed the benchmark-as-code import and merged an MCP path for a new account’s agent to create its benchmark.", "selection_reason": "Omitted because the architecture choice is settled and your subsequent hands-on benchmark thread already establishes familiarity with the resulting workflow." }, { "meeting_id": "GPM16-02", "substance": "Person D’s open reply-format PR reports worked-example gains, an eval/live-context mismatch and safeguards against cutting necessary information.", "selection_reason": "Omitted in favor of the distinct diagnostic-count experiment; the afternoon event here is approval of a September 15 head, not a newly established experiment." }, { "meeting_id": "GPM16-03", "substance": "Engineer A’s open retry PR documents degraded-success handling and regressions where suppressing diagnostic counts also removed useful timings.", "selection_reason": "Selected the reported prompt interaction as a reusable learning, retaining the small samples and pending final evaluation rather than presenting a shipped fix." }, { "meeting_id": "GPM16-04", "substance": "Person P’s merged external-agent profile change separates self-reported capabilities from observed handoff behavior and marks unobserved profiles explicitly.", "selection_reason": "Selected the concise claim-versus-evidence rule and its concrete implementation without claiming deployment or proven routing improvements." }, { "meeting_id": "GPM16-05", "substance": "Baby Agent’s merged external-employee feature can brief another Slack bot and resume after its answer; review fixed a case that discarded replies after roster changes.", "selection_reason": "Omitted as a separate card because the related profile rule in GPM16-04 is the stronger reading moment, and real-model and production validation remain limited." } ], "notes": "Read all 25 supplied primary units in full, including selected GitHub patches and later thread corrections. Notion transcripts lack speaker labels and contain mixed turns and transcription errors; the selected meeting attributions remain visibly tentative. Slack authors and GitHub accounts are explicit, but PR prose and reported experiments may be AI-assisted and were not independently reproduced. Supplied source URLs and original L-locators are retained; linked images, videos and outside pages were not inspected. Used only the supplied dated strategy reference and historical dedup/calibration material, with no assumption of absence or novelty from retrieval. No external messages, publication or product actions were performed." }