Chinh (lelouvincx) / Weekly Report - 2026-W10

Created Mon, 25 May 2026 00:00:00 +0000 Modified Mon, 25 May 2026 06:02:25 +0000
1386 Words
page

Weekly Report — Mar 4 – Mar 7, 2026

Pull Request: #6

Friday comprehensive report for 2026-W10. Activity concentrated Feb 25 – Mar 3 (journal entries), with PR review follow-ups continuing Mar 4–6.

  • Work

    Add2Cart

  • (done) Fix the temp bridge table — eliminated bridge_product_retailer by adding retailer_id directly to atc_price_history via slug normalization. Reduced join path from 4 tables to 2–3 for price queries. Backfilled 30M+ rows: 23/40 retailers matched cleanly (40% of rows), 6 near-misses identified for follow-up. Stored procedures updated; sp_run_all reordered. Holistics AML dataset relationships updated to use direct retailer_current_pricing_skey join. (Notion)

  • (done) Document all stored procedures in a data lineage — reviewed 13 stored procedures across 3 architecture layers (source transforms → derived facts → aggregates). Created Notion KB with refresh patterns, validation logic, scheduling recommendations, and Mermaid lineage diagram. SPs are currently not version-controlled and have no scheduler. Verdict: keep SPs over dbt migration due to low ROI. (Notion)

  • (done) Fix small issue in dashboard 3 (Promotional Planning) — filter was not excluding products outside the primary retailer (18 vs 17 product count mismatch). (Slack)

  • (done) Catch up with Simon on Add2Cart progress. (ReadAI)

    Internal

  • (in-progress) Review MRR PR (holistics/dbt#812) — deep code review on customer identity resolution refactor (DAT-471). The PR decomposes itg_mappings__customer_identities into 5 sub-models. Found 2 critical issues: (1) Zoho-only rows silently dropped by WHERE clause, (2) NULL domains filtered by NOT IN semantics. Also identified Zendesk fan-out risk affecting 3 active paying tenants. Submitted “changes requested” review. Review discussion continued Mar 4–6 with author (hellofromtheothersky) resolving comments. Also flagged: update relationships.dbml for new models, linting fixes needed, CI test failure on unrelated model (stg_holistics__jobs).

  • (in-progress) Answer Vincent & Sriram on trial-first vs call-first conversion data — created DE-202 Notion doc. Initial analysis of won deals: ~81% trial-only, ~19% trial+demo call. Key limitation: current data only captures Calendly demo calls, not all sales touchpoints. Small deal count magnifies mis-categorization (e.g., MrDIY). Next step: share POC with Vincent and Sriram before full build. Draft AML explore on chinh/lead-funnel-by-sales-motion branch.

  • (in-progress) Continue commenting DE-201 (PR Workflow enhancement) — carry-over. Audit revealed internal-aml-project repo has zero CI/CD, dbt CI +1 hop limit misses critical singular tests 2–3 hops away. Proposed tiered roadmap: (1) quick wins — update PR templates, add CODEOWNERS; (2) expand test coverage — tag critical singular tests, implement dbt model contracts; (3) new automation — lineage diff, impact analysis, data-diff tools. (Notion)

  • (done) Answer Quinn on tracking admin edits to tenant subscriptions. (Slack)

    Presales

  • Highlight: Trust Wallet win story — won over from Looker/Tableau. Crypto wallet app on Databricks stack; high-touch onboarding was key differentiator: live enablement on Model → Dataset → Dashboard flow, tackling struct fields, typing, and caching for Databricks cost concerns.

  • Note: Vincent’s demo call with Mangrove Digital (experienced DAX/LookML practitioners) — new sales narrative to pre-empt Omni’s spreadsheet formula appeal. Covers objection handling on dbt metrics integration, AQL learning curve, metric promotion, and AI pricing.

  • (done) Onboarding call with Fluent Data. (Clari)

    Duty Support

  • (done) Share AI usage data for Sweetspot — completed Feb 27. (Slack)

  • Personal / Tooling

  • (done) Scheduled weekly Logseq report automation — launchd cron with Tuesday + Friday runs. Shell script + amp CLI headless agent. (GitHub PR#3)

  • (done) Set up git credentials for lelouvincx-bot — separated SSH/signing keys from 1Password; configured PAT for gh CLI. (Gemini)

  • (done) Set up Notion CLI skill for Amp. Completed Feb 26.

  • (done) Set up Holistics query skill for Amp — enables querying Holistics datasets via Semantic API. (GitHub PR)

  • (done) Set up BigQuery CLI skill for Amp. (Ampcode)

  • (done) Set up Claude Code — explored as comparison to Amp CLI. Verdict: Amp’s tool calling is noticeably better.

  • Learning & Notes

  • “Down The Rabbit Hole” — self-reflection on recurring patterns of over-engineering and yak shaving. Identified symptoms: spending hours perfecting things that could be simpler, side-questing into dotfile/neovim debugging when main task is dashboard work, “just one more small fix” loops.

  • Impressions from Vincent’s demo call (Benjamin Franklin writing practice) — practiced analytical writing by reviewing a recorded demo. Key observations: (1) talking continuously is a skill worth practicing; (2) hitting the spot when answering questions matters most for impression; (3) knowing even small product features (e.g., AI field descriptions) creates “I’m annoyed I haven’t tried this” moments; (4) be honest about limitations — don’t overestimate features (e.g., AI pricing transparency). (ReadAI)

  • Benn Stancil: “No Really, Everything Becomes BI” — insight is relative; it depends on what people already know. You can’t tell people something surprising without knowing what is expected. Insight isn’t the data, it’s what’s not in the audience’s head. (Source)

  • Benn Stancil: “Go Crazy Folks, Go Crazy” — recurring pattern in software: incumbents are afraid to ship bold changes, then startups do it anyway (Google vs ChatGPT, Copilot vs Claude Code). Art parallel: when art becomes easy to make, value drops — but craft becomes extremely valuable. (Source)

  • Amp CLI: “200k tokens is plenty” — agents degrade with too many tokens (“get drunk”). Use threads strategically: each small task as a new thread, use read_thread to transfer context. (Source)

  • Google Data Portability API — noted for potential data export use cases.

  • Idea: Archive all Amp threads into a repository as backup against context loss (like ChatGPT history issues).

  • Idea: Personal weekly report automation — designed the full workflow concept: branch-per-week, diff-based note extraction, enrichment from Slack/Notion/GitHub/Ampcode, headless amp CLI execution via launchd.

  • agents.craft.do/mermaid — beautified Mermaid rendering as a JS package, potentially useful for docs/websites.

  • Next Week

    From pages/Backlog.md, ordered by priority:

    P1

  • (none explicitly P1 in backlog — current P1 items are in-progress or done)

    P2 — Should Do

  • Internal: Fix excluding internal testing Zoho accounts (DAT-524) — lingering since week 7

  • Internal: Continue Cross-Model Calculation — timing may still be premature (docs revamping not far enough)

  • Add2Cart: Document Add2Cart data model; product feedback to team (2 items)

  • Presales: Add tabs to Hooli dashboard — write solution proposal (Slack)

  • Duty Support: Document Medenterprises performance use case — overdue since Feb 5

  • Teaching: Figure out set-based thinking curriculum; find football dataset for Duc Anh; checkout pudding.cool resources; research visualization resources

  • Personal: Consider a Flashcard app for learning English words

    P3 — Nice to Have

  • Presales: Read Modeling Patterns docs; watch Taxdome onboarding call; learn embedded analytics use case; compare ChatGPT solutions Omni vs Holistics

  • Duty Support: Check with Product on metric hover detail; document Tricia’s A/B comparison use case

  • Personal: Setup linear CLI agent skill; self-host readeck; migrate remaining Obsidian notes

    Carry-over from this week (in-progress)

  • Review MRR PR holistics/dbt#812 — awaiting author’s remaining fixes

  • DE-202 trial-first vs call-first — share POC with Vincent/Sriram

  • DE-201 PR Workflow enhancement — continue commenting

  • Career & Personal Consulting

    Observations from this week:

    1. High-impact code review work: The MRR PR review (#812) was substantive — finding 2 critical data integrity issues (silent row drops, NULL domain semantics) demonstrates strong analytical ability. This kind of deep review is high-leverage: it prevents production data quality issues that would be expensive to debug later. Continue investing in thorough reviews; it builds trust and domain expertise.

    2. Over-engineering awareness is a growth signal: The “Down The Rabbit Hole” self-reflection shows mature self-awareness. The pattern of yak-shaving (dashboard task → dotfile debugging → neovim → 3 hours lost) is common in senior engineers. You’ve identified it — now consider a concrete countermeasure: time-boxing side-quests to 15 minutes, or keeping a “rabbit holes” list to revisit during designated personal time.

    3. Tooling investment is paying off: This week saw 4 agent skills set up (Notion, Holistics, BigQuery, weekly report automation). The weekly report automation alone will save 1–2 hours per week. The risk is spending more time building tools than using them — consider a cooling-off period before starting new skill builds.

    4. Breadth across projects: You touched Add2Cart, Internal (2 projects), Presales, Duty Support, Teaching, and Personal tooling in one week. This breadth is valuable but watch for context-switching costs. The week plan (via Ampcode thread) shows good prioritization discipline — keep using it.

    5. Writing practice: Using Benjamin Franklin’s method to analyze Vincent’s demo call is an underrated investment. Sales engineering and presales work require clear communication; this practice compounds over time.

    6. Backlog hygiene: Several items have been lingering (DAT-524 since week 7, Medenterprises since Feb 5, Cross-Model Calculation since Nov). Consider either re-prioritizing these or explicitly deferring them to reduce mental overhead.