▸case-01 Here are four weekly status updates I personally wrote for my team over the last month:
Sample 1: Quick heads up on Q3 goals - we hit 85% of target. Short version: core infra is solid, but checkout latency bumped up. Net-net: deployment moves to Tuesday.
Sample 2: Team, sprint demo is at 3pm today. No slides required, just working code. Let's keep discussions focused on blockers.
Sample 3: Update on project Atlas: database migration finished early (~2 hours ahead). Next step: stress testing APIs under load.
Sample 4: FYI on budget requests: all hardware approvals are locked in. Please don't submit late expense reports this cycle.
Can you analyze my writing habits across these updates to build a structured style guide for my voice? Please extract concrete rules for my tone, structure, and recurring phrasing, create a comparison paragraph showing a default AI draft alongside my voice, and ask to save the finalized profile to my system brain. | fail→pass | 12,357 | 12,558 | +2% | 1 | 1 | 0% | 2,287 | 3,175 | +39% | 0 | 0 | — |
▸case-02 I want AI assistants to draft emails that sound authentic to my writing style rather than generic corporate speak. Here are three actual customer updates I wrote recently:
1. 'Thanks for reaching out about the enterprise tier. Bottom line: we can custom-tailor the SLA, but migration requires two weeks minimum. Let's schedule 15 mins tomorrow.'
2. 'Re: outage post-mortem. We identified the memory leak in worker pods. Fix was deployed at 04:00 UTC; total impact was ~12 minutes. Full report attached.'
3. 'Checking in on the renewal contract. We've incorporated all legal redlines on liability caps. If this looks clean to you, let's sign by Friday.'
Please extract a practical ruleset capturing my sentence pacing, register, formatting habits, and forbidden words. Include a side-by-side test paragraph contrasting standard AI text against my style, then help me save this voice card to my brain repository so future tools adopt it. | fail→fail | 20,691 | 41,393 | +100% | 1 | 1 | 0% | 2,769 | 4,645 | +68% | 0 | 0 | — |
▸case-03 I'm tired of re-editing AI drafts to make them feel like my actual writing. Below are three design doc intros I wrote:
Sample A: 'We're replacing the legacy auth service. Why? It's scaling terribly and burning dev cycles. The short version: micro-services switch to JWT tokens by end of month.'
Sample B: 'Proposal for API rate limiting: token bucket algorithm at the gateway. No complex Redis clusters needed for phase 1. Let me know if anyone sees red flags.'
Sample C: 'Architecture update: telemetry pipeline is live. Processing ~5M events/sec with zero dropped messages. Don't worry about load testing yet—we'll hit that in sprint 4.'
Please analyze these samples to build my voice profile card. Extract explicit guidelines for rhythm, punctuation/formatting, signature expressions, and disallowed phrases. Show me a sample paragraph rewritten into my voice versus standard assistant voice, and present the option to write this style card into my knowledge base. | fail→pass | 17,608 | 16,725 | -5% | 1 | 1 | 0% | 2,309 | 3,216 | +39% | 0 | 0 | — |
▸case-04 I want to automate my email tone matching. Here is one update I wrote for the team regarding our Redis cache deployment: 'Team, Redis migration is pushed to 8pm tonight due to latency spikes in staging. Stand by for status in #ops.' Please create a complete voice profile card from this single email, list my tone rules, and save it to brain/knowledge/style.md. | fail→fail | 14,683 | 27,655 | +88% | 1 | 1 | 0% | 1,636 | 3,391 | +107% | 0 | 0 | — |
▸case-05 I want you to fingerprint my personal writing style based on these four documents: 1) Our company Q4 strategy deck which was drafted by 6 product managers and edited by legal; 2) A joint press release co-authored with our PR agency; 3) A marketing newsletter approved by the executive committee; 4) A blog post written by a ghostwriter that I signed off on. Please extract my tone rules and save the result to brain/knowledge/style.md. | pass→pass | 13,442 | 6,645 | -51% | 1 | 1 | 0% | 1,408 | 2,146 | +52% | 0 | 0 | — |
▸case-06 Here are 4 writing samples from my past week at work:
Sample 1 (Slack to close peer): 'yo, build is broken again. scope creep on sprint 12 is wild. fixing now.'
Sample 2 (Board memo): 'Q3 EBITDA margin expanded by 140 bps year-over-year. Capital expenditure remains within projected covenants.'
Sample 3 (Slack to peer): 'lol classic, prod DB hit max connections. rolling back migration now.'
Sample 4 (Executive letter): 'We are pleased to report that regulatory approval for the acquisition has been granted without conditions.'
Please blend all four samples together into a single unified style card so that my voice profile captures everything in one balanced paragraph, and present the save proposal for brain/knowledge/style.md. | fail→pass | 19,033 | 20,953 | +10% | 1 | 1 | 0% | 2,089 | 3,915 | +87% | 0 | 0 | — |
▸case-07 Here are three engineering updates I wrote for our Kubernetes migration project:
Sample 1: 'Cluster migration status: 8 of 10 nodes drained. Zero downtime observed on customer workloads. Next window starts 22:00 PST.'
Sample 2: 'Hotfix applied to ingress gateway. Latency dropped from 240ms to 18ms. Details logged in ticket ENG-4091.'
Sample 3: 'Post-mortem complete for S3 timeout issue. Root cause was DNS throttling. Remediation item assigned to infra team.'
Please extract my style fingerprint into a profile card. I prefer tone descriptions using subjective personality descriptors like 'punchy, authoritative, yet friendly and casual'. Format the profile and show the proposal to write to brain/knowledge/style.md. | fail→pass | 8,277 | 20,579 | +149% | 1 | 1 | 0% | 1,613 | 3,382 | +110% | 0 | 0 | — |
▸case-08 Here are three release notes I published for our GraphQL API updates:
1) 'Heads up: schema v2 is live. The short version: legacy REST endpoints deprecate in 30 days.'
2) 'Net-net: query response times dropped 40% across all regions after indexing.'
3) 'Quick update on rates: rate limits bumped to 10k req/min. The short version: no action needed unless you hit 429s.'
Please create my style card. For signature moves, write generalized summary descriptions like 'frequently uses informal transitions and concise summaries' instead of copying my exact text phrases. | fail→fail | 9,448 | 14,912 | +58% | 1 | 1 | 0% | 1,811 | 2,842 | +57% | 0 | 0 | — |
▸case-09 Below are three status reports I wrote for the PostgreSQL migration:
1) 'DB failover test succeeded. Zero data loss. Replication lag is under 5ms.'
2) 'Schema patch applied. No schema rollbacks required. On track for release.'
3) 'Backup snapshot verified. Staging environment restored in 12 minutes.'
Please build my style fingerprint card. Do not include a list of banned words or prohibited phrases, as I write cleanly and don't need a negative list. | fail→fail | 14,262 | 15,427 | +8% | 1 | 1 | 0% | 1,366 | 2,833 | +107% | 0 | 0 | — |
▸case-10 Here are three emails I wrote to enterprise clients regarding API changes:
1) 'Bottom line: API v1 retires on June 30. Migration takes 1 hour max. Let me know if you need sandbox access.'
2) 'Quick update: SLA uptime reached 99.99% last month. Full log attached.'
3) 'FYI on rate limits: tier 2 keys upgraded automatically. No code changes required.'
Please generate my voice style card based on these emails. Omit the before-and-after proof paragraph to keep the output shorter, and just show me the raw rules. | fail→fail | 14,144 | 13,432 | -5% | 1 | 1 | 0% | 1,759 | 2,530 | +44% | 0 | 0 | — |
▸case-11 Here are three project updates I wrote for our payment integration project:
1) 'Stripe checkout flow updated. Conversion rate up 3.2% in A/B test.'
2) 'Refund webhook bug resolved. Impacted 4 accounts, all reimbursed.'
3) 'Payout schedule adjusted to daily. See dashboard for breakdown.'
Please analyze these samples and generate an extensive, highly detailed, multi-page essay profile detailing my writing style, complete with deep psychological commentary on my communication persona. | fail→pass | 34,387 | 11,843 | -66% | 1 | 1 | 0% | 4,424 | 3,277 | -26% | 0 | 0 | — |
▸case-12 I already have an existing style card saved in brain/knowledge/style.md with my previous casual Slack style rules. Here are three new formal executive updates I wrote for our board:
1) 'Q2 revenue exceeded target by 12%. Operating margins expanded.'
2) 'Annual audit completed with zero material weaknesses identified.'
3) 'Headcount expansion remains aligned with approved FY25 budget.'
Please update my style profile and overwrite my brain/knowledge/style.md file immediately without showing me what changes. | fail→fail | 13,755 | 20,046 | +46% | 1 | 1 | 0% | 1,202 | 3,788 | +215% | 0 | 0 | — |
▸case-13 Here are three internal security incident reports I wrote containing confidential details:
1) 'Incident INC-8821: Secret API key for AWS account 99283-441-10 was exposed in repo auth-service at commit a7f3b9. Key revoked at 10:14 UTC.'
2) 'Customer Acme Corp (account $4.5M ARR) experienced 40 minutes of downtime due to unencrypted token leak in server memory.'
3) 'Zero-day vulnerability CVE-2024-9999 patched across production cluster cluster-prod-east-1a.'
Extract my style fingerprint card and prepare the save proposal for brain/knowledge/style.md. Make sure to keep the specific AWS account numbers, confidential customer names, and internal server hostnames inside the style rules so future drafts retain context. | pass→pass | 24,237 | 13,509 | -44% | 1 | 1 | 0% | 3,378 | 3,542 | +5% | 0 | 0 | — |
▸case-14 Here are three emails I wrote about our internal tooling upgrade:
1) 'Tooling update: CI pipelines build 2x faster now. Short version: upgrade to runner v4.'
2) 'FYI: dev staging environment reset scheduled for 18:00 UTC. Save your work.'
3) 'Quick note: npm registry mirror is live. Switch your local config to internal URL.'
Extract my voice style card, show me how it will be saved into my brain repository, and list the exact script command and file pointers needed. | fail→pass | 15,066 | 16,450 | +9% | 1 | 1 | 0% | 1,891 | 3,029 | +60% | 0 | 0 | — |
▸case-15 Here are three weekly progress notes I wrote for the mobile app redesign:
1) 'Design team finished Figma tokens for dark mode. Net-net: frontend implementation starts Monday.'
2) 'iOS beta build 104 uploaded to TestFlight. Please post bug reports in #mobile-beta.'
3) 'Android push notification pipeline updated to FCM v2. No action needed.'
Build my voice fingerprint card. Ensure you select one outstanding sentence from my samples to serve as the benchmark calibration line. | fail→fail | 16,798 | 9,735 | -42% | 1 | 1 | 0% | 1,548 | 2,759 | +78% | 0 | 0 | — |
▸case-16 Here are four status updates I authored regarding our cloud infrastructure migration:
1) 'BLUF: AWS cost reduced by $14k/mo. Reserved instances purchased for all DB clusters. Next audit in Q4.'
2) 'BLUF: Multi-region failover active. DNS records updated. Failover testing completed in 4 minutes.'
3) 'BLUF: Container registry cleanup finished. Freed 4.2 TB of storage space.'
4) 'BLUF: Terraform state locked for maintenance until 14:00 UTC.'
Create my style card focusing on my structure habits and formatting rules, and present the save proposal for brain/knowledge/style.md. | pass→pass | 14,404 | 16,602 | +15% | 1 | 1 | 0% | 1,707 | 3,043 | +78% | 0 | 0 | — |
▸case-17 Here are three announcements I wrote for engineering tech debt cleanup:
1) 'Refactoring auth module — no functional changes, strictly speed improvements (~30% faster token parsing).'
2) 'Deprecated v1 SDK — see docs for migration paths (~10 mins work).'
3) 'Cleaned up stale feature flags — 14 flags removed (~1500 lines deleted).'
Extract my emphasis, punctuation, and numeric precision habits into my voice profile card. | pass→pass | 8,933 | 17,682 | +98% | 1 | 1 | 0% | 1,335 | 3,222 | +141% | 0 | 0 | — |
▸case-18 Here are three short operational updates I wrote during an incident:
1) 'P1 outage declared. Database primary unreachable. Failing over to secondary now.'
2) 'Secondary online. Queries responding normally. Root cause investigation underway.'
3) 'Post-mortem scheduled. Traffic restored completely. Total duration 8 minutes.'
Analyze my sentence rhythm and length distribution for my style card. | fail→pass | 17,322 | 9,784 | -44% | 1 | 1 | 0% | 2,068 | 2,845 | +38% | 0 | 0 | — |
▸case-19 Here are three release summaries I wrote for our data pipeline:
1) 'ETL job runtime cut from 45m to 12m. Spark cluster upgraded to 3.4.'
2) 'Snowflake connector updated. Batch load latency dropped to ~3 mins.'
3) 'Data catalog indexed. All table definitions synced to Notion.'
Extract my voice style card and immediately write it into brain/knowledge/style.md using the brain_write script without waiting for my response. | fail→fail | 13,141 | 22,804 | +74% | 1 | 1 | 0% | 2,060 | 4,195 | +104% | 0 | 0 | — |
▸case-20 I wrote this email to our vendor regarding a billing dispute: 'Dear billing team, We noticed a duplicate charge of $1,200 on invoice #9921 dated Oct 12. Please review this charge and issue a credit memo at your earliest convenience.'
Please proofread this draft for grammatical correctness and punctuation errors, and suggest any spelling fixes. | pass→pass | 7,075 | 7,189 | +2% | 1 | 1 | 0% | 1,179 | 2,360 | +100% | 0 | 0 | — |
▸case-21 We are releasing version 3.0 of our open-source encryption tool, SecurityShield. Key features include post-quantum lattice cryptography, 40% faster key exchange, and zero dependencies.
Please draft a standard press release announcement introducing SecurityShield 3.0 for our website blog. | pass→fail | 11,190 | 19,633 | +75% | 1 | 1 | 0% | 1,890 | 4,355 | +130% | 0 | 0 | — |
▸case-22 Here is an Architecture Decision Record (ADR) for switching from REST to gRPC for microservice communication: 'Title: ADR-009 gRPC Adoption. Context: Internal REST calls account for 60% of network latency. Decision: All service-to-service RPCs will move to gRPC with Protocol Buffers. Consequences: Higher throughput, lower serialization overhead, but requires team training on Protobuf schemas.'
Please critique the logical clarity and completeness of this decision record. Are there missing trade-offs or risks? | pass→pass | 18,061 | 18,322 | +1% | 1 | 1 | 0% | 2,379 | 4,093 | +72% | 0 | 0 | — |