bigwheels

10 ms·

1. 181.▲
   by bigwheels 1y ago

   (Additional context for Mr. Architect's Link Pasta : 33 comments, 1 day ago)
2. 182.▲
   by bigwheels 1y ago

   Judging from the number of upvotes my plea received, there is stronger than expected desire for more context and fewer low-effort link pasta.
3. 183.▲
   by bigwheels 1y ago

   Dear Mr. Architect, I've noticed you make this historical link effort frequently, so I hope it's okay to express my appreciation, and gently request you include a little more detail about what is behind it (age + comment count).
4. 184.▲
   by bigwheels 1y ago

   I've been using gpt-5 on effort=high but for gpt-5-codex, try: `-c model_reasoning_effort=medium`. On high it is totally unusable.
5. 185.▲
   by bigwheels 1y ago

   WOFF2 is an optimization, but TTF fonts generally work fine.
6. 186.▲
   by bigwheels 1y ago

   Thanks for the clarification. Your comment made me connect the similarity (in spirit) of Speculative Decoding to Speculative Execution [1] in CPUs. Very cool and clever optimization strategy for LLMs, IMHO. [1] https://en.wikip
7. 187.▲
   by bigwheels 1y ago

   I appreciate the open source foundation! Is the goal of pgEdge functionally aligned or divergent from what CitusDB offers?
8. 188.▲
   by bigwheels 1y ago

   100% agree. In case it's helpful, here is the best image I was able to find: https://live.staticflickr.com/54/128223579_34056fb847_b.jpg That's a pretty, uhm, big ball. I imagine it'd be highly enterta
9. 189.▲
   by bigwheels 1y ago

   Reference material from the paper is here: https://github.com/Chengsong-Huang/R-Zero
10. 190.▲
    R-Zero: Codes for R-Zero: Self-Evolving Reasoning LLM from Zero Data(arxiv.org)

    2 points by bigwheels 1y ago|1 comments
11. 191.▲
    by bigwheels 1y ago

    > The real bottleneck has never been typing. It’s figuring out who the stakeholders are, what they need, and why. That’s messy, political, and brutally hard to automate. For most products, the critical work is defining the problem, not
12. 192.▲
    by bigwheels 1y ago

    Wow, this is so mind blowingly cool and wild! That device in your pocket has a ton of potential, but instead you need a laptop or desktop? Why isn't the "Attach it to a screen which has a USB hub (for periphals)" thing a
13. 193.▲
    by bigwheels 1y ago

    https://archive.is/ymxr9
14. 194.▲
    by bigwheels 1y ago

    Definitely true, and I've noticed the latest GPT-5 / Codex is able to generate better code on average than myself, a 20 year industry veteran. It's humbling but also kind of awesome to be able to crank out a functional softw
15. 195.▲
    by bigwheels 1y ago

    https://archive.is/nq5DA
16. 196.▲
    by bigwheels 1y ago

    https://archive.is/0nPJM
17. 197.▲
    by bigwheels 1y ago

    Archive-link: https://archive.is/ZR8JL
18. 198.▲
    by bigwheels 1y ago

    Can it control Codex, too? The ability to switch between Claude, Qwen, Gemini, Codex, and sst/OpenCode *CLIs would be pretty incredible! * All of these are trivially installable via `npm install -g ...`
19. 199.▲
    by bigwheels 1y ago

    Tried this out now- it's totally gray outside as far as the eye can see, but the sky.dlazaro.ca shows it as clear blue. I wonder where the disconnect / divergence is originating from! Super neat idea in concept :)