How Slashy Reviews Code That Runs in Your Inbox
When your code lives in someone's inbox, a bug isn't a log line. It's a sent email.
Background
Slashy builds an AI agent that plugs into the apps people already work in, Gmail, Slack, Notion, JIRA, calendars, and CRMs, and takes real action inside them: drafting and sending email, booking meetings, opening tickets, updating records. Founded in 2025 by Pranjali Awasthi, Dhruv Roongta, and Harsha Gaddipati, and backed by Y Combinator, it's used by founders, investors, and sales teams to hand off the repetitive work.
That sets a high bar for the code. Slashy doesn't just render information, it acts on it, in your real accounts, with your credentials. A normal bug shows something wrong on a screen. A bug here sends the wrong email or updates the wrong record. It doesn't crash. It acts. So the team is unusually serious about code review.
A review stack measured on the scoreboard
Slashy doesn't take a review tool's word for its value. It runs several in parallel on the same pull requests and scores each one against GitHub's own records: did an engineer resolve the comment, or change the exact code it flagged? That is what Slashy counts as "actioned," and it is the only metric that matters, because it separates the comments that moved the code from the ones that got scrolled past.
Over three months across Slashy's two repositories, engineers opened roughly 2,400 pull requests. Entelligence reviewed 870+ of them, leaving 2,300+ inline review threads.
The result
The signal is in the trend. In the most recent month measured, engineers actioned 87% of Entelligence's comments, resolving them or shipping the exact change they suggested. In a stack running multiple review bots side by side on the same code, that put Entelligence level with the strongest reviewer in the stack and ahead of the rest. It is Entelligence's best month on record at Slashy, and every figure is reproducible from GitHub's own data, not an internal dashboard.
The numbers
All from GitHub's own review-thread records, across Slashy's two repos over the three-month window:
87% of Entelligence's comments accepted in the latest month, up from 61% across the full window
2,400+ pull requests opened in the window
870+ PRs reviewed by Entelligence, about 1 in 3 of every PR opened
2,300+ inline review comments left, roughly 2.7 per reviewed PR
54% of comments resolved by an engineer, 37% applied as the exact change in a later commit
Why 87% is the number that matters
On a codebase that acts in users' real accounts, a review comment isn't a style nit you can wave off. It's a wrong email you haven't sent yet. Nearly nine in ten of Entelligence's comments driving a real code change means the team is treating those findings as what they are: the last check before the code goes and does something on someone's behalf.
Bottom line
Slashy holds its review tools to GitHub's own scoreboard, not to vendor claims. Measured that way, in the latest month, 87% of Entelligence's comments were acted on, level with the best reviewer in a crowded stack, on a codebase where a missed bug doesn't crash. It hits send.