Miso vs Muse

From the reports

What we wrote when Miso and Muse took the same test. Every update here has its own page you can link to.

1. Oct 13, 2026Speed weighting and smoother mobile transitions- Booking speed now accounts for 40% of Travel; the other seven dimensions share 60%. Individual reviewed scores, timings and evidence are unchanged. - The resulting scores are Miso 9.3, Instinct 8.0, Grok Bot 7.9, Dots 7.1 and Muse 6.9. The report, Travel list/grid, profiles and comparison use the shared calculation. Missing dimensions remain unscored; observed weights are renormalized and coverage adjustment is retained. - Booking speed evidence and timing scope remain distinct from the report’s search-through-checkout ten-minute assessment. - Scroll-driven effects remain on mobile. Settled and hidden chapters avoid redundant updates; title and scene transforms are applied directly to their layers. Resize measurements are batched, and browser-toolbar changes skip unnecessary layout work. - The split-flap board keeps its visual motion with fewer text elements. Rescheduling planes use transforms for continuous movement. Mobile report headers use opaque white rather than blurring moving content underneath. - Regression checks cover weighting, missing data, reverse scrolling, hash jumps and chapter accessibility. Browser-size checks supplement rather than replace testing on physical phones.Read the updateAI Assistant Consumer Travel Report: October 2026
2. Oct 13, 2026From AI Assistant Consumer Travel Report: October 2026- Strong planning. - No confirmed seat or completed rebooking. - Not especially fast; not the full-workflow choice.Read it in the report
3. Oct 12, 2026Travel results reorganized as a visual recap- Seven capability questions replace the article structure. - Booking approaches include overlapping browser and API tools. - The ten-minute threshold covers search, preferences and checkout, based on the reviewer’s assessment. - Seat positions are illustrative; descriptions summarize confirmed seat-upgrade actions and required help. - Reviewed benchmark scores and recorded run evidence are unchanged. - A tinted opening header ends at the author; the named, illustrated contents bar follows it and becomes sticky while scrolling. - A compact goal statement and four-stage visual replace the two introductory requirement cards. - Three boarding-ticket cards show the booking methods, with assistant logo stamps and a one-time stack-to-spread reveal. - Replay buttons, chapter counters, and duplicate captions are removed; outcome summaries use direct wording. - A centered scroll stage now connects all nine chapters; local animation progress follows scrolling in either direction. - An airport split-flap board replaces the booking rows and reveals that all five assistants booked successfully. - Compact visuals and a shorter verdict keep each chapter within the stage. Reduced motion retains a readable static layout. - A longer, left-facing plane uses a booking-style seat layout; assistant action summaries appear above it without cabin or exact-position claims. - Seat-action descriptions follow the reviewed seat-selection results. - The rescheduling scene removes its duplicate summary and route labels. Planes continue moving only while the chapter is active. - Grok’s rescheduling note specifies the checkout stall and the traveler’s click on “Change flight,” following the reviewed change results. - A split-open airplane reveals a Basic Economy fare card with four assistant stamps; Miso remains on a separate stub. The visual describes fare availability, not a cabin division or a fare purchase. - Five airport service desks replace the cancellation rows and crossed route lines. Scroll-driven screens confirm all five cancellations; brief notes retain browser help and refund outcomes from the reviewed cancellation results. Bank settlement remains unverified. - The traveler-information scene now follows a transcript audit of all ten supplied chat exports for the five report assistants. All five collected or reused traveler details; Miso, Dots and Muse initiated airline-loyalty requests; no unprompted Known Traveler Number request appears in the supplied chats. - TSA-style checkpoints separate basic detail handling from unprompted requests. Miso’s use of saved PreCheck information is identified separately. The audit records source locations in `data/travel-profile-request-audit.json`; it contains no private message text or identity values. Reviewed profile and proactiveness scores remain unchanged. - The speed chapter spreads the two under-ten-minute assistants and three over-ten-minute assistants around a central stopwatch. A repeating rim sweep runs while the chapter is active, retaining the fixed ten-minute threshold and the reviewer’s full-flow assessment. Reduced motion keeps a static comparison. - The verdict is titled “Best overall.” Miso’s lack of Basic Economy is presented as a major budget limitation; Instinct is the fast Basic Economy alternative, with its Duffel rescheduling restriction visible. This practical alternative does not change the reviewed overall ranking. - A full table follows the scroll journey, with all five assistants side by side, observed trip outcomes, separately audited information requests, and all eight reviewed dimension scores. A “Table” contents link reaches this ordinary scrolling recap. - Round two removes the paper receipt cards from the cancellation desks, retaining the stands, cancellation screens and small refund-verification note. - Removed the bank-settlement footnote from the closing comparison table. - The opening header adds a faint globe grid, three slow flight routes and subtle destination pings. Decorative motion is hidden from assistive technology and stops for reduced-motion preferences. - A fixed, quiet mode toggle switches between the animated journey and a plain article with the same findings and comparison table. Only the selected view is mounted, and switching retains the current chapter. - Hero flight routes extend past both screen edges so planes enter and leave off-screen. - The Known Traveler Number checkpoint now includes Miso’s unprompted app-onboarding collection, supported by the accompanying report and the user’s clarification. None asked for the number over text; the article and recap table distinguish app collection from transcript requests. - The hero background removes the globe grid, route lines and destination pings, leaving six small planes moving slowly across the clean tinted header. - Removed the opening subtitle and the seat, fare, cancellation, profile-summary and speed captions requested in the six screenshots. - Mobile audit replaces the wide closing table with screen-width trip comparisons and a five-column score matrix whose assistant labels remain visible while scrolling. Phone scenes use a stable small viewport height, retain full-size text and 44px touch targets, and switch to normal flow when a pinned chapter would not fit. A separate bottom control strip prevents the mode toggle covering content. Short-phone and landscape mode switching restores the matching chapter; supporting text is enlarged and off-screen hero flights pause. - Mobile scrolling animations are restored on short and landscape phones. The top contents bar becomes a bottom drop-up alongside Article mode, with equal centered buttons. Taller visuals pan within their chapter as scroll progress advances, keeping native text size and a visible heading. The desktop contents bar and mobile comparison remain intact. - The closing mobile comparison is replaced with a single link to the existing Travel score grid, in both viewing modes. The desktop comparison remains unchanged.Read the updateAI Assistant Consumer Travel Report: October 2026