Skip to content

Commit b90968e

Browse files
committed
Update README to clarify context-first approach, emphasize accuracy and simplicity, and refine language for better clarity
1 parent 84d6465 commit b90968e

1 file changed

Lines changed: 4 additions & 4 deletions

File tree

‎README.md‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -19,16 +19,16 @@ Context-first AI coding for VS Code, Cursor, and others. Free, open-source, and
1919

2020
## Introduction
2121

22-
Large language models are brilliant, yet mindless pattern matchers. They can't by themselves _imagine_ what noise (examples) could help them generate better outputs. To mitigate this limitation, other tools ask you to put effort into planning—a practice of expanding prompts, so the model can do better job at finding relevant files. This is slow, token-hungry, and dilutes model's attention for the task.
22+
Large language models are brilliant, yet mindless pattern matchers. As they lack any mental models—they can't _imagine_ how the correct output should look like ahead of selecting files for context. To mitigate this limitation, other tools ask you to put effort into planning—a practice of expanding instructions, so the model can do a better job at finding relevant files. This is slow, token-hungry, and dilutes model's attention for the task.
2323

24-
Code Web Chat places context ownership in developer's hands. By guiding the model with hand-picked files you'll **save time** by writing simple instructions, **save money** by being token-efficient, ensure **privacy** by never sending your codebase for external indexing and **security** by not running arbitrary commands.
24+
Code Web Chat places context ownership in developer's hands. Guide the model with hand-picked files and get **the best accuracy** by not polluting the context window with tool definitions and endless rambling, **save time** by writing simpler instructions (goodbye to markdown files), **save money** by being token-efficient, ensure **privacy** by not having to send your whole codebase for external indexing and **security** by not running arbitrary commands.
2525

2626
**Context-first approach**
2727

28-
Meet non-agentic workflow—select files that set the environment for the task, enter instructions, and send message with your favorite chatbot or a model provider of choice.
28+
Meet a simple, non-agentic workflow—select files, enter instructions, and send message with your favorite chatbot or a model provider of choice.
2929

3030
> [!TIP]
31-
> LLMs are pattern matchers—they love examples! Steer the model to higher quality outputs, like no coding agent can.
31+
> LLMs are pattern matchers—they love examples! Guide the model with context to production-grade code.
3232
3333
Constructed prompt is straightforward and focus the model's whole attention on the task. Example:
3434

0 commit comments

Comments
 (0)