- Analyze 16 posts at a time by default, up from 4. A run of 50 posts takes
about half the time.
analysis.concurrency still accepts 1 through 16.
- Add the
texts input. The Actor analyzes your own texts & fetches nothing
from X for them. Each analyzed text costs the same as an analyzed tweet.
- Run every source of a mixed input: lists, tweet IDs & searches share
maxItems. Earlier the first source ran alone.
- Accept 16 more field names from other tweet Actors, such as
sort,
tweetLanguage, author, start & end.
- Add the Algorithm Score estimate. The Actor applies the ranking weights X
published to each post's public likes, replies, reposts & quotes. Every row
gains
viral.algorithmScore, viral.algorithmBasis,
viral.algorithmWeightedSum, viral.actualEngagementRate &
viral.weightsVersion (x_algorithm_params:2026-09-18).
- Add the flat columns
viralAlgorithmScore & viralActualEngagementRate to
every row & to the Viral Score dataset view.
- Add
viral.calibration to the run summary. It compares Viral Scores with
actual engagement rates & lists up to 5 overperformers & 5 underperformers.
- Add
viral.accounts to the run summary: averages, best & worst post & bucket
averages for up to 50 handles.
- Add
viral.leaderboard to the run summary: the top 20 handles by average
Viral Score & by average actual engagement rate.
- The new numbers come from data each row already carries. The Actor asks the AI
no new question, & the AI never sees engagement counts.
- Keep every existing output field unchanged.
- Add the
viral_lite default questions: hook, clarity, informative, funny,
ragebait, AI-written style, spam risk & likely crowd reaction.
- Add the Viral Score estimate from 0 to 100 with a
send_it, edit_first or
sleep_on_it verdict. Every row carries viral, viralScore &
viralVerdict (weights version viral_lite:1).
- Cap the score with hard stops for spam, ragebait & machine-like style.
- Report the average score & verdict counts in the
analysis-summary record &
the run report.
- Add the
Viral Score dataset view for spreadsheet exports.
- Reuse filtered extraction, run comparison & durable successful-analysis
billing.
- Add input, result & diagnostic schemas.
Live qualification & publication remain pending.