tweedler290
Joined 2 months ago
Karma
3
Submissions
22
Member since
Aug 20
Recent submissions
Comment · 16 days ago
Comment · 1 month ago
honestly with computer vision so OP now QR and barcodes should be dead. like here's a (probable) easy fix: make a hella contrasty square 1x1. location of a single dot inside it determines the product. the precision of locating that dot reliably? train a vision model hella hard
Comment · 1 month ago
thats crazy. i made this https://github.com/oooscoos/Benzi which does the same thing but is language agnostic (supports 9 other languages). urs probs goes in deeper into c++ than mine tho
Comment · 1 month ago
GitHub: https://github.com/oooscoos/Benzi
Demo: https://varianttech.net/demo
Benchmarks: https://varianttech.net/benchmark
Roughly speaking, the way current AI coding agents/harnesses work is by either:
a) Pulling in appropriate text snippets of code across multiple files and handing them to the agent, or
b) Parsing code to make high dimenional embeddings to approximate a symptom map, and hand that to the agent.
Both of these approaches skyrocket the token count, add to wall clock time, contribute to context drifting, add to the model's thinking tokens to discover the structure of the program, and then FORGET most of it when Claude Code compacts, or ALL of it if it's a multifile refactoring because all line numbers shift and need re-grepping.
Benzi is built from the ground up to AVOID reading source code in the first place. It supplies the artificial intelligence model deterministic intelligence via tool calls. For example, when a model is about to make a code change, it could query "what functions feed this one?" -- half the time it isn't even necessary because the Benzi compiler already informs it of the blast radius before and after making edits, along with a complete static analysis check.
Benzi Sonnet reads far less source code (9,125 lines) than Claude Code Sonnet (20,704), DeepSeek's harness (43,598), and OpenCode (65K+ LOC -- disqualified due to repeated failure) to accomplish the same tasks faster and cheaper. (benchmark link in comments)
Benzi has truth tiers clearly seperating what can be analyzed with static analysis from what can't -- and then adding a runtime tracer on top to bridge the gap between the two (details in FAQ on github).
It also has several bonus features such as a runtime tracer, syntax & semantic verified writes, context aware model written repro, mid task model upgrade if task is too difficult, and SEVERAL more.
It currently supports Python · JavaScript · TypeScript · Java · C# · C++ · C · Go · Rust · Ruby, and can handle HTML, CSS and JS -- deterministically. Claude Code clicks photos, Benzi resolves winners of CSS rules. The CodeIndex and the MarkupIndex are fairly well tested, and if something isn't working, the model is made aware of it first.
On the benchmarks side, 78.2% SWE-bench Verified for <10¢ a fix (using V4flash). This score is noteable because while the rest of the industry is leaning plugin-heavy and pouring millions of dollars into increasing context window sizes, Benzi's approach might prove to be economically more valuable while improving the model's code writing/comprehenion abilities.
Thanks for reading! please let me know what you think. I am aware the AI fatigue is real, but I hope you can see why indexing a repo > reading raw source code. Please go over the github readme before snap judgements..
Comment · 1 month ago
GitHub: https://github.com/oooscoos/Benzi
Demo: https://varianttech.net/demo
Benchmarks: https://varianttech.net/benchmark
Roughly speaking, the way current AI coding agents/harnesses work is by either:
a) Pulling in appropriate text snippets of code across multiple files and handing them to the agent, or
b) Parsing code to make high dimenional embeddings to approximate a symptom map, and hand that to the agent.
Both of these approaches skyrocket the token count, add to wall clock time, contribute to context drifting, add to the model's thinking tokens to discover the structure of the program, and then FORGET most of it when Claude Code compacts, or ALL of it if it's a multifile refactoring because all line numbers shift and need re-grepping.
Benzi is built from the ground up to AVOID reading source code in the first place. It supplies the artificial intelligence model deterministic intelligence via tool calls. For example, when a model is about to make a code change, it could query "what functions feed this one?" -- half the time it isn't even necessary because the Benzi compiler already informs it of the blast radius before and after making edits, along with a complete static analysis check.
Benzi Sonnet reads far less source code (9,125 lines) than Claude Code Sonnet (20,704), DeepSeek's harness (43,598), and OpenCode (65K+ LOC -- disqualified due to repeated failure) to accomplish the same tasks faster and cheaper. (benchmark link in comments)
Benzi has truth tiers clearly seperating what can be analyzed with static analysis from what can't -- and then adding a runtime tracer on top to bridge the gap between the two (details in FAQ on github).
It also has several bonus features such as a runtime tracer, self-aware model upgrade mid task if it thinks the job is over its pay grade, context aware model written repro, and SEVERAL more.
It currently supports Python · JavaScript · TypeScript · Java · C# · C++ · C · Go · Rust · Ruby, and can handle HTML, CSS and JS -- deterministically. Claude Code clicks photos, Benzi resolves winners of CSS rules. The CodeIndex and the MarkupIndex are fairly well tested, and if something isn't working, the model is made aware of it first.
On the benchmarks side, 78.2% SWE-bench Verified for <10¢ a fix (using V4flash). This score is noteable because while the rest of the industry is leaning plugin-heavy and pouring millions of dollars into increasing context window sizes, Benzi's approach might prove to be economically more valuable while improving the model's code writing/comprehenion abilities.
Thanks for reading! please let me know what you think. I am aware the AI fatigue is real, but I hope you can see why indexing a repo > reading raw source code. Please go over the github readme before snap judgements..
Try feeding the live demo any random ahh repo you want. And ask it questions only real code intelligence could answer.
Comment · 1 month ago
ur most welcome! if you're curious how I found it, I have made a code intelligence system. I literally fed it your repo and asked for bugs. You can try it yourself here: https://benzi.fly.dev/
Comment · 1 month ago
Roughly speaking, the way current AI coding agents/harnesses work is by either: a) Pulling in appropriate text snippets of code across multiple files and handing them to the agent, or b) Parsing code to make high dimensional embeddings to approximate a symptom map, and hand that to the agent.
Both of these approaches skyrocket the token count, add to wall clock time, contribute to context drifting, add to the model's thinking tokens to discover the structure of the program, and then FORGET most of it when Claude Code compacts, or ALL of it if it's a multifile refactoring because all line numbers shift and need re-grepping.
Benzi is built from the ground up to AVOID reading source code in the first place. It supplies the artificial intelligence model deterministic intelligence via tool calls. For example, when a model is about to make a code change, it could query "what functions feed this one?" -- half the time it isn't even necessary because the Benzi compiler already informs it of the blast radius before and after making edits, along with a complete static analysis check.
Benzi Sonnet reads far less source code (9,125 lines) than Claude Code Sonnet (20,704), DeepSeek's harness (43,598), and OpenCode (65K+ LOC -- disqualified due to repeated failure) to accomplish the same tasks faster and cheaper. (Benchmark details: https://benzi.fly.dev/benchmark)
Truth tiers are maintained with clear and transparent boundaries. Explained in the github readme's FAQ. (https://github.com/oooscoos/Benzi)
It also has several bonus features such as a runtime tracer, self-aware model upgrade mid task if it thinks the job is over its pay grade, context aware model written repro, and SEVERAL more.
It currently supports Python · JavaScript · TypeScript · Java · C# · C++ · C · Go · Rust · Ruby, and can handle HTML, CSS and JS -- deterministically. Claude Code clicks photos, Benzi resolves winners of CSS rules. The CodeIndex and the MarkupIndex are fairly well tested, and if something isn't working, the model is made aware of it first.
On the benchmarks side, 78.2% SWE-bench Verified for <10¢ a fix (using V4flash). This score is notable because while the rest of the industry is leaning plugin-heavy and pouring millions of dollars into increasing context window sizes, Benzi's approach might prove to be economically more valuable while improving the model's code writing/comprehenion abilities.
If you're curious to learn more, click https://benzi.fly.dev/about
Comment · 1 month ago
Token cost and wall clock time are practical pain points for SWEs, but the bigger intellectual pin point is that the dang bot has to reconstruct the code structure actually READING the code as TEXT.
like do we NOT KNOW what functions are? how they could change variables? or what in an "if/else" would do? Now think about a compiler, does it not ALREADY DO ALL THAT? just it compiles to machine code - not something an AI can use. Benzi is built to bridge that gap. Fully language agnostic and model agnostic, it compiles the deterministic intellgience for instant O(1) lookups, so your artifical intellgience can do the real SWE work. Tree-sitter is the only dependency.
Please let me know what you think and star on GitHub if you like it
Comment · 1 month ago
Token cost and wall clock time are practical pain points for SWEs, but the bigger intellectual pin point is that the dang bot has to reconstruct the code structure actually READING the code as TEXT.
like do we NOT KNOW what functions are? how they could change variables? or what in an "if/else" would do? Now think about a compiler, does it not ALREADY DO ALL THAT? just it compiles to machine code - not something an AI can use. Benzi is built to bridge that gap. Fully language agnostic and model agnostic, it compiles the deterministic intellgience for instant O(1) lookups, so your artifical intellgience can do the real SWE work. Tree-sitter is the only dependency.
Please let me know what you think and star on GitHub if you like it
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