Senslop

Inspiration

I never liked the code review experience in the IDEs I tried. A normal diff felt overwhelming, especially when a coding agent changed several related files at once. I built Senslop because I wanted a calmer way to understand what changed and a review workspace I actually wanted to use.

What it does

Senslop is a local code review workspace for developers who want to understand changes and catch bugs before pushing. Install the CLI with your favorite package manager, run senslop app start, and review in a local browser workspace.

• Capture working-tree changes, a commit, a range, or a branch comparison.
• Group related code into a suggested reading order so definitions and dependencies come before their consumers.
• Inspect diffs and captured source, take notes, mark regions complete, and reopen a review later.
• With a configured model, ask questions about the code, explain a selection, or run a correctness review with findings linked to original source evidence.
• Decide which findings matter, hand off a fix task, inspect returned edits, run a targeted check, and compare review rounds.

Model assessments, check results, and human decisions stay separate.

How we built it

The CLI starts a local Next.js and React browser app plus a background job engine. Git supplies source changes, SQLite stores review history, and TypeScript contracts connect the CLI, app, and review core. AI features use a provider I configure; reading a diff works without AI.

I used the Devpost Learn Skill Pack to record the idea, scope, PRD, technical spec, and build checklist. Codex helped implement and verify the workflow as it grew from reading changes to investigating findings and checking fixes.

Challenges we ran into

The working tree keeps changing, but a review must still explain the exact code it originally captured. Senslop saves source and evidence per review round and keeps model judgments, passing checks, and human decisions distinct. That prevents a newer file or a confident model answer from being mistaken for proof that an older issue was fixed.

Accomplishments that we're proud of

• Built a local workflow from capturing a change through reading, investigating a finding, inspecting a fix, running a check, and comparing the next round.
• Built an application I will actually use in my day-to-day code review workflow.
• Added automated tests, local report export, and an optional Git push checkpoint.

What we learned

Building Senslop reinforced that a useful AI review needs a clear reading path and inspectable evidence, not just a model verdict. Keeping the source snapshot, checks, and my decisions separate made the review easier to trust and revisit. Planning with the Learn Skill Pack helped keep the expanding workflow tied to an explicit scope and spec.

What's next for Senslop

I want to keep refining the reading experience and validate more models and operating systems. I also want to make the CLI installation and first-run experience as simple as the review workflow itself.

Built With

• git
• inngest
• next.js
• node.js
• react
• sqlite
• typescript
• vercel-ai-sdk