swiftlyTyped

7 ms·

1. 61.▲
   RAG vs. Context-Window in GPT-4: accuracy, cost, & latency(ai88.substack.com)

   18 points by swiftlyTyped 3y ago|8 comments
2. 62.▲
   by swiftlyTyped 3y ago

   TL;DR: (RAG + GPT-4) delivers superior performance, at 4% of the cost.
3. 63.▲
   by swiftlyTyped 3y ago

   Looks cool. What do you mean by cursor-style insertions/edits?
4. 64.▲
   Show HN: CopilotTextarea: <Textarea/> W GPT Completions Like Gmail/GitHub. OSS(github.com)

   5 points by swiftlyTyped 3y ago|0 comments
5. 65.▲
   by swiftlyTyped 3y ago

   Copilots are considered unsexy bc they appear not to aim at AGI (i.e. go for full autonomy). But that gets the game theory completely backwards. LLMs (and agents) can often do 90% of the heavy lifting, but those 10% failures (or just weak r
6. 66.▲
   Show HN: CopilotKit – a hackable OSS copilot for any react app(github.com)

   10 points by swiftlyTyped 3y ago|2 comments