Cameron Keith

AI research and D1 Golf @Dartmouth | building @BramaAI

[Cameron Keith @CamKeithGolf]

Oops Opus 5, 53.4 < 53.5... What happened to the agent that was supposed to audit the table after creating it?

[Claude @claudeai]

On several coding and knowledge work evaluations, Opus 5 is the new state-of-the-art:

[Cameron Keith @CamKeithGolf]

Both

[Kelvin Celso @kelvinbuildss]

Are you team Claude fable 5 or Codex 5.6 sol?

[Cameron Keith @CamKeithGolf]

Guess Fable is here to stay?!

[Mira Murati @miramurati]

Our first model, Inkling. Trained from scratch, weights are open, fine-tunable on Tinker today.

[Cameron Keith @CamKeithGolf]

Anthropic is running a masterclass in artificial scarcity with their models

[elvis @omarsar0]

I don't think Anthropic realizes how disruptive these changes are to users. I appreciate the extension, but please stop playing games. Either keep it under the subscriptions or put it under the API already.

[Cameron Keith @CamKeithGolf]

@superset_sh is really solid

[Marty Kausas @marty_kausas]

i'm so sick of using claude code in a terminal

i'm not coding. who has made a great app that i can use with any model?

[Cameron Keith @CamKeithGolf]

Let's go!

[ClaudeDevs @ClaudeDevs]

We've reset 5-hour and weekly rate limits for all users.

[Cameron Keith @CamKeithGolf]

A "global workspace" is a real cognitive-science concept, not marketing. If it holds up, the interesting part isn't that Claude thinks silently, it's that we can now point to where. Interpretability is quietly becoming the an important AI subfield.

[Polymarket @Polymarket]

BREAKING: Anthropic researchers reveal they found a “global workspace” inside Claude that allows it to think silently.

[Cameron Keith @CamKeithGolf]

If it also became a 100x plan

[Justin Schroeder @jpschroeder]

Would you upgrade to a $1000/month Claude plan to keep Fable?

[Cameron Keith @CamKeithGolf]

@claudeai Fable is cool, but when is the 100x plan dropping?

[Cameron Keith @CamKeithGolf]

@sridharfyi

[Cameron Keith @CamKeithGolf]

So clutch

[Elon Musk @elonmusk]

Cursor for iOS!

[Cameron Keith @CamKeithGolf]

Perf lives in the software stack, not the silicon. Baseten just raised $1.5B building a company around exactly this... squeezing more out of the same NVIDIA chips everyone else has. Benchmarks measure ceilings; the stack decides what you actually hit.

[rewind @rewind02]

AI researcher breaks down why your $5,000 Mac Studio runs local AI slower than benchmarks promise:

"Performance isn't about the hardware - it's about three software layers stacked on top of it."

in 15 minutes he tests MLX, Ollama, llama.cpp, and the new vllm-mlx head to head

no GPU needed, just the Mac you probably already own

- why Ollama, the most popular tool, is actually one of the slower options on Mac
- the real bottleneck nobody benchmarks: 'prefill' time before the first token even appears
- the exact quantization setting (4-bit) that's the best speed-to-quality tradeoff
- why unified memory makes Macs uniquely good for big models compared to Nvidia GPUs

complete guide to choosing the right hardware for local AI is below ����

[Cameron Keith @CamKeithGolf]

Let’s see if @cursor_ai can survive until my @claudeai weekly usage resets 😂

[Cameron Keith @CamKeithGolf]

Thanks @ycombinator and @cursor_ai for the YC event credits… can’t wait to token max my cursor account too #StartupGrowth

[Cameron Keith @CamKeithGolf]

The comparison cuts both ways. Dot-com had revenue projections that never showed up; semis today have real earnings and real demand. The risk isn’t fake growth, it’s that even real growth gets priced past perfection. ‘Extremes tend to rhyme’ is the right framing.

[Charlie Bilello @charliebilello]

Semiconductor stocks are now up 246% over the last 14 months, surpassing the 234% surge during the peak of the dot-com bubble.

The AI mania has taken semis to a place only seen once before.

History doesn’t repeat, but extremes tend to rhyme.

Video: https://t.co/WStwCcfM0J