Everything completed since late May, week by week, in plain English — pulled from the actual work record.
9 weeks~105 updates shipped2 engine rebuilds6+ new user-facing features
Each item is tagged by purpose:Scoring engineNew capabilitiesSimplicityFoundations
Week of May 25
~22 updates
Scoring engineBegan rebuilding the heart of our proprietary scoring engine from the ground up — more reliable, self-contained, and built to be defensible.
Scoring engineTaught the engine to gather and weigh real supporting evidence, so every score rests on proof instead of opinion.
SimplicityRebuilt the score results screen — a clearer score display with an 18-factor breakdown — and retired the old cluttered pages.
Week of June 1
~9 updates
New capabilitiesSplit the product into two clear journeys — Brand Research and Campaign Evaluation — so research done once can be reused for every later campaign instead of starting over.
SimplicityAdded a "what this score means" panel that explains every score in plain language.
Scoring engineTightened the rules so the engine can only cite legitimate sources — it can never treat its own opinions as evidence.
Week of June 8
~13 updates
FoundationsBroke the backend's biggest tangle of code into clean, focused parts — verified to behave identically, and much safer to build on.
FoundationsReplaced aging plumbing between analysis steps, fixing bugs where re-runs could quietly use stale information and produce weaker results.
Week of June 15
~16 updates
New capabilitiesBuilt the first version of the Insights Showcase — a designed, presentation-ready report layout for findings. (The full visual overview ships Aug 4.)
FoundationsFinished the backend cleanup and moved to our AI vendor's new platform ahead of a forced deprecation, avoiding outages.
Week of June 22
~8 updates
Scoring engineStarted the summer's biggest rebuild: an "evidence ledger" that logs every piece of proof as the research runs, making every score transparent and auditable.
Scoring engineFindings are now ranked by how strong their evidence is — not just how often something gets mentioned.
Week of June 29
~9 updates
Scoring engineEvery strategic concept now gets its own calibrated score and rating band, not just one number for the whole run.
Scoring engineUpgraded the AI "judge" that reviews close calls, with guardrails that flag any disagreement between the AI and the math for audit.
Scoring engineChecked the engine against reality: a famously successful real-world campaign scored near the top of the scale, and a famous flop scored near the bottom.
Week of July 6
~23 updates
Scoring engineSwitched the new evidence-based scoring system on end to end, retired the old one, and safely migrated every existing analysis to it.
Scoring engineEvery scored claim now traces back to the actual web sources behind it — with those sources shown right on the score page.
New capabilitiesFirst live run of concept generation: behind the scenes, the engine can now create three campaign concepts in different creative styles and score each one.
SimplicityMerged a major interface overhaul and added a short executive summary to every section of the report.
FoundationsA reliability push: clear error messages with a retry button, graceful recovery from dropped connections, and always-on error monitoring.
Week of July 13
~3 updates
SimplicityThe plain-language executive summary now refreshes itself automatically after every run — it can never be out of date.
SimplicityShipped the Audio Overview: a two-minute, two-host podcast recap of any analysis, with an in-app player, transcript, and one-click regenerate.
FoundationsConsolidated the summer's scoring rebuild and partner contributions into the main product line.
Week of July 20
In progress
SimplicityReplaced a confusing mid-run decision step with a simple risk slider — set how bold or safe you want the strategy, and the analysis runs straight through on its own.