Changelog
What's new in Hackabacus
Every update, improvement, and fix shipped to the Hackabacus Excel and PowerPoint plugin. Follow along as we build the best finance toolkit available.
Preserve Named Styles in PowerPoint Tables
- Excel tables exported to PowerPoint now preserve fonts, fills and borders from named cell styles, including Heading 1, Comma, firm templates and Hackabacus style cycling
- Fixed cells losing formatting when an Excel number format contains a greater-than sign
Fast, Formatted Excel Tables in PowerPoint
- Paste linked Excel ranges as editable PowerPoint tables quickly while preserving source fonts, colors, borders, displayed values and merges
- Accounting formats and international text now transfer without stray characters
- Refresh uses the same fast formatting path and keeps the existing table if an update fails
PowerPoint Chart Fill Selection
- Select a chart series with the mouse to color all its columns, or click again into one column to color only that point
- Palette colors, the main Fill Color button, and No Fill now follow the same series or point selection
- Applying a color to a whole series also replaces individual column colors and restores columns previously set to No Fill
File Versioning & Cloud PDF Export Fixes
- Save Up now advances beyond the highest matching file version, handles padded suffixes, and ignores unrelated files with similar names
- PDF to Folder now saves beside local, OneDrive, and SharePoint presentations; only unsaved presentations fall back to Documents
- Copied HTML content now uses explicit fragment markers for more reliable clipboard offsets
Deterministic AI Review Tie-Outs
- AI Review now inventories native PowerPoint text, tables, cached chart values, and Hackabacus-linked sources locally to tie EBITDA, IRR, MOIC, equity need, revenue, valuation, leverage, and other labeled metrics across slides
- Confirmed and possible inconsistencies are separated while scenario, period, scope, currency, adjustment basis, gross/net, and entry/exit differences remain distinct
- Equivalent displays such as $1.2bn and $1,200mm normalize before comparison, with rounding tolerance based on displayed precision
- The model receives only compact conflict candidates in one review request, materially reducing token usage while retaining local tie-outs if the provider fails
- Local OpenAI-compatible models gain private-network validation, model discovery, structured-output fallback, and clearer privacy and context-size warnings
Want to see what's coming next?
Check out our product roadmap for upcoming features, or let us know what you'd like to see in the next release.
View roadmap