quantisan

they want our interaction data

I'm surprised Stanford Law would go along with this over-reaching press release title. How about "For common first-year contracts-law questions, law professors preferred AI-generated answers to professor-generated answers"

Does this mean they're effectively a PE firm?

Try Cmd-Shift-J. It dumps your buffer to a temp file, with the file path made ready for you. I do that and then open the temp file with vi.

Blog post https://qwen.ai/blog?id=4074cca80393150c248e508aa62983f9cb7d...

I use Pico with LLM code gen for new projects. As you probably know, LLMs are predisposed to Tailwind and coding for industrial strength on any tiniest projects. The trick is to feed it the whole Pico docs as context and prompt it (i.e. in your CLAUDE.md) to use Pico explicitly.

1. Zero-Copy Deserialization with rkyv and Unsafe Code 2. Platform-Specific High-Performance Memory Allocators 3. High-Performance Hash Maps with FxHasher and SwissTable

https://deepwiki.com/search/point-to-the-top-3-rust-specif_4...

I used Claude Code to understand the source code of Claude Code:

  1. Tool-Based Action Framework
    - Tools implement a common interface (Tool.ts) with standard methods
    - Central registry in tools.ts manages tool availability and permissions
    - Tools are categorized as read-only or stateful, affecting execution strategy
  2. Agent Orchestration
    - query.ts serves as the orchestration engine that:
        - Identifies tool use requests in Claude's responses
      - Manages permissions and serializes/deserializes tool data
      - Schedules concurrent or serial tool execution based on tool type
      - Feeds tool results back to Claude in a recursive loop
  3. AgentTool Implementation
    - Enables recursive agent capabilities via sub-agent spawning
    - Sub-agents run with controlled permissions (typically read-only)
    - Each agent invocation is stateless and returns a single result
    - Implements progress streaming via async generators
  4. Execution Flow
    - User query → Claude response → Tool use requests → Permission checks
    - Tool execution → Results normalized → Results fed back to Claude
    - Process repeats with Claude potentially using additional tools
  5. Architectural Patterns
    - Async generators for streaming results and maintaining responsiveness
    - Hierarchical permission model controlling tool access
    - Normalized message formats ensuring consistent communication
    - Logging chains for tracking agent activities
    - Context management with optional compression (/compact)

- Youth-Led Implementation - Bypassing Traditional Authority Structures - Loyalty-Based Appointments - Institutional Disruption - Limited Oversight - Information Control

https://en.wikipedia.org/wiki/Cultural_Revolution

"I'll never buy a SBC from Nvidia unless all the SW support is up-streamed to Linux kernel," top comment from prev discussion https://news.ycombinator.com/item?id=42623030

As others have said, this came after historically significant low popularity and mounting political pressures. His government faced criticism over falling poll numbers, by-election losses, and a broken agreement with the NDP. Tensions with the US and internal dissent within the Liberal party added to his challenges.

One of the final nails in the coffin was the resignation of Chrystia Freeland, his last standing ally and Finance Minister.

This video from CBC a couple weeks ago on Freeland explains the rifts in Trudeau's government well https://www.youtube.com/watch?v=8SuZTLWNlpc

Previous years:

2024: https://news.ycombinator.com/item?id=38777115 (the rest of this list is copied from https://news.ycombinator.com/item?id=38779963)

2023: https://news.ycombinator.com/item?id=34125628

2022: https://news.ycombinator.com/item?id=29746236

2021: https://news.ycombinator.com/item?id=25594068

2020: https://news.ycombinator.com/item?id=21802596

2019: https://news.ycombinator.com/item?id=18753859

2018: https://news.ycombinator.com/item?id=16007988

2017: none?

2016: https://news.ycombinator.com/item?id=10809767

2015: https://news.ycombinator.com/item?id=8822723

2014: https://news.ycombinator.com/item?id=6994370

2013: none?

2012: https://news.ycombinator.com/item?id=3395201

2011: https://news.ycombinator.com/item?id=1970023

2010: https://news.ycombinator.com/item?id=1025681

thanks for sharing

Amazing! OpenAI figured out how to scale inference. https://arxiv.org/abs/2407.21787 show how using more compute during inference can outperform much larger models in tasks like math problems

I wonder how do they decide when to stop these Chain of Thought for each query? As anyone that played with agents can attest, LLMs can talk with themselves forever.

it's still available https://balsamiq.com/wireframes/desktop/

Balsamiq offers a desktop version with a one-time license fee at $149 per user https://balsamiq.com/wireframes/desktop/

Copy edited their blurb:

Core is an experimental tool that simplifies the process of creating Action Role-Playing Games (Action-RPGs) by providing a unified set of tools and a streamlined approach to game development. It represents game elements (entities) and their properties (components) as simple data structures, allowing for greater flexibility and easier modification. The entire game state is stored in a single container (app/state), making it easier to manage and update the game's data. Core also provides a graphical user interface (GUI) for editing game content stored in a single file (resources/properties.edn), making the development process more accessible to non-programmers. By using Malli schemas for data validation, Core ensures that the game content is consistent and error-free, reducing development time and improving the overall quality of the game.

Rick and Morty ref?

https://youtu.be/rYJP2YWtZl8?si=NIvyeVwCJaExq92U

for some stories on Clojure in production, check out https://www.juxt.pro/clojure-in/

https://en.wikipedia.org/wiki/The_Myst_Reader

Price (input) $0.35 / 1 million tokens (for prompts up to 128K tokens) $0.70 / 1 million tokens (for prompts longer than 128K)

Price (output) $0.53 / 1 million tokens (for prompts up to 128K tokens) $1.05 / 1 million tokens (for prompts longer than 128K)

---

Compared to GPT-3.5 Turbo

Input US$0.50 / 1M tokens Output US$1.50 / 1M tokens

I had multiple Microsoft Lumia as my primary driver for a couple years. I've been on iPhones for 10+ years total. There were some good (better user interface) and bad (Live logins required in random places) with Windows Mobile. Like Windows, they have pockets of good ideas fractured by their lack of cohesion.

I worked with Fiber Bragg grating [1] technology in the 2000s, exploring its applications in diverse fields. This included biomedical applications, where a fibre optic cable inserted inside a catheter measured temperature and pressure at specific points within a person's body. And in deep oil drilling, where the same fibre optic cable could extend kilometers down a well.

[1] https://en.wikipedia.org/wiki/Fiber_Bragg_grating

twinflights is something I could use. Is it possible to select number of stop overs on each direction? e.g. often I like to wander on way out from home. But then on way back I just want to get home ASAP.

In terms of the UX, a table view is a bit daunting while I'm starting with an exploration mindset, perhaps start the user with a map instead?

Also FYI, I randomly clicked on Book a flight. It led me to Kiwi but the price went up from 158 to 198. I tried a few other buttons and the price stayed the same.

Thanks for sharing this, looking forward to see where you take it.

thanks for sharing the human aspect to this story! I had to fact check it. For others wondering, https://en.wikipedia.org/wiki/Spirited_Away some mentioning in second paragraph.

Apple's own LLM and major Siri update is supposed to be coming in iOS 18 next year. So says the rumour, https://www.macrumors.com/2023/09/06/ios-18-siri-improvement...

this was fun. Any suggestions for other casual co-op games on the phone that I can play with my almost-4 year old?

https://jalammar.github.io/visualizing-neural-machine-transl... and https://jalammar.github.io/illustrated-transformer/ for anyone looking

this one's been mentioned a lot: Let's build GPT: from scratch, in code, spelled out. https://youtu.be/kCc8FmEb1nY

It sounds like they have a stack that's working already, is that correct? If that's the case, why are you tasked to re-design it? Why not keep things running and improve only when and where it's needed?