HackerReposA FIELD GUIDE TO OPEN SOURCE
ProSponsor the shelf♧ 0
← Back to top all-time
oooscoos
Benzi
HTML2026–2026
oooscoos / Benzi

Benzi

Benzi is code intelligence infrastructure for AI models. It compiles your codebase into a resolved, queryable map of calls, data flow, control flow and class hierarchy first; then it reads, explains, runs, edits and verifies your code using it. Every language runs its own tree-sitter grammar into the same query map.

3Distinct people
14Mentions
0Explicit recommenders
177GitHub stars
REPOSITORY AT A GLANCE

What to know before exploring

What it is

Benzi is code intelligence infrastructure for AI models. It compiles your codebase into a resolved, queryable map of calls, data flow, control flow and class hierarchy first; then it reads, explains, runs, edits and verifies your code using it. Every language runs its own tree-sitter grammar into the same query map.

GitHub

Where it is used

No documented product use has been verified yet.

What people are saying

No recommendation phrase was found in the indexed comments.

No classified warning is attached to this page. Read the discussions before treating silence as approval.

Also in these discussions

No other repository is repeatedly linked in the same discussions.

This week on Hacker News

Nov 9–Nov 16 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/16/2026, 8:59:02 AM UTC.

No direct repository links were found in the indexed HN discussions over the past seven days. Older discussions are available below.

Mentions over time

HN mentions by week14 dated mentions

Observed mentions: 2026-09-05 19:22:25 UTC to 2026-10-17 16:22:10 UTC. Weeks start Monday, UTC.

0123452026-08-31 00:00:00 UTC to 2026-09-07 00:00:00 UTC (exclusive): 1 mentions; 1 submissions, 0 commentsAug 312026-09-07 00:00:00 UTC to 2026-09-14 00:00:00 UTC (exclusive): 2 mentions; 1 submissions, 1 comments2026-09-14 00:00:00 UTC to 2026-09-21 00:00:00 UTC (exclusive): 5 mentions; 1 submissions, 4 comments2026-09-21 00:00:00 UTC to 2026-09-28 00:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 commentsSep 212026-09-28 00:00:00 UTC to 2026-10-05 00:00:00 UTC (exclusive): 3 mentions; 0 submissions, 3 comments2026-10-05 00:00:00 UTC to 2026-10-12 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2026-10-12 00:00:00 UTC to 2026-10-19 00:00:00 UTC (exclusive): 2 mentions; 0 submissions, 2 commentsOct 12

Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.

View mention counts and date ranges
UTC ranges include their start and exclude their end. Select a period to filter the source evidence.
PeriodStart (UTC)End (exclusive, UTC)MentionsSubmissionsComments
Aug 312026-08-31 00:00:00 UTC2026-09-07 00:00:00 UTC110
Sep 72026-09-07 00:00:00 UTC2026-09-14 00:00:00 UTC211
Sep 142026-09-14 00:00:00 UTC2026-09-21 00:00:00 UTC514
Sep 212026-09-21 00:00:00 UTC2026-09-28 00:00:00 UTC101
Sep 282026-09-28 00:00:00 UTC2026-10-05 00:00:00 UTC303
Oct 52026-10-05 00:00:00 UTC2026-10-12 00:00:00 UTC000
Oct 122026-10-12 00:00:00 UTC2026-10-19 00:00:00 UTC202

What Hacker News says

2 sources
Benzi is a code intelligence system. With the agentic loop + compiler it scores 82.4% on SWE-bench Verified for $12.96 while reading 2-4x less source code than other agents/harnesses (measured separately, check 24-bug benchmark). Code intelligence itself is old, but bringing AI into the mix changes things. The AI models generate TEXT. The written text happens to be code for SWE purposes. Now, what improves with every model iteration is the cognition/comprehension of whatever text it has generated, in our case, code. Anthropic intrinsically doesn't care if its models chew through the codebase raw, their business is cognition improvement. For codebases with millions of lines of code that require a lot of thinking tokens regardless of what model is used, they rely on model cognition, not code intelligence. Benzi is compiler backed. It suppor
Read the rest
ts 10 programming + 3 frontend languages (and is model agnostic - not loyal to any frontier lab). In fact, the whole tier taxonomy of RESOLVED/CANDIDATE/OBSERVED ( + external/unresolved/unindexed) was built around resolving the repo once, and letting the agent query that, while updating incrementally on edits and automatically reverting broken ones. The runtime tracer is how an edge gets OBSERVED in the first place. The compiler is lean and compiles locally. It is currently available for free as a VS code ext. and CLI. The Benzi compiler can also be queryed by other agents - the MCP server exposes only the compiler, your agentic loop stays with your agent. github: https://github.com/oooscoos/Benzi (benzi is not OSS. repo contains swe bench results + code benzi wrote - you should check it out, the compiler changes how model designs changes). website is there too but I dont want to make this look like a promo. really what i want is for you guys to just try it out once. the demo loads any random repo within seconds - throw your own at it even. https://varianttech.net/demo
In [dead]
Benzi is a code intelligence system. With the agentic loop + compiler it scores 82.4% on SWE-bench Verified for $12.96 while reading 2-4x less source code than other agents/harnesses (measured separately, check 24-bug benchmark). Code intelligence itself is old, but bringing AI into the mix changes things. The AI models generate TEXT. The written text happens to be code for SWE purposes. Now, what improves with every model iteration is the cognition/comprehension of whatever text it has generated, in our case, code. Anthropic intrinsically doesn't care if its models chew through the codebase raw, their business is cognition improvement. For codebases with millions of lines of code that require a lot of thinking tokens regardless of what model is used, they rely on model cognition, not code intelligence. Benzi is compiler backed. It suppor
Read the rest
ts 10 programming + 3 frontend languages (and is model agnostic - not loyal to any frontier lab). In fact, the whole tier taxonomy of RESOLVED/CANDIDATE/OBSERVED ( + external/unresolved/unindexed) was built around resolving the repo once, and letting the agent query that, while updating incrementally on edits and automatically reverting broken ones. The runtime tracer is how an edge gets OBSERVED in the first place. The compiler is lean and compiles locally. It is currently available for free as a VS code ext. and CLI. The Benzi compiler can also be queryed by other agents - the MCP server exposes only the compiler, your agentic loop stays with your agent. github: https://github.com/oooscoos/Benzi (benzi is not OSS. repo contains swe bench results + code benzi wrote - you should check it out, the compiler changes how model designs changes). website is there too but I dont want to make this look like a promo. really what i want is for you guys to just try it out once. the demo loads any random repo within seconds - throw your own at it even. https://varianttech.net/demo
In [dead]

MAKE YOURSELF AT HOME

A few shortcuts

/
Search the collection
?
Open this guide
Esc
Close dialogs and navigation

Your saved repositories and interests stay in this browser.