Predictions stay accurate through a stock split

If an assistant tracked a stock prediction against a company that later did a stock split, the comparison could come out wrong — the original price it recorded didn’t reflect the split, but the current price always does, so the two weren’t measuring the same thing anymore.

Assistants that score predictions now re-check the price on a consistent basis at scoring time, so a split no longer throws off the result.

A market data outage now says so

When Ward’s stock and market data source had an outage, tools could return an empty result that looked the same as “nothing to report” — no error, no way to tell the difference from your assistant’s output.

Market tools now say when the data source itself is unavailable, instead of quietly returning nothing.

More reliable reading of major news feeds

Assistants that follow the news — headline briefings, market watchers — now read BBC and Financial Times feeds directly and more reliably. Reuters and AP don’t offer a public feed for automated readers to follow, so assistants fall back to searching those instead of retrying a feed that was never going to work.