Tom Boyle @tomlikestocode

【 º ᗜ º 】founder @swiftcompute | swe, ai, ml | bb 🥋 | making compute cheaper swiftcompute.ai San Francisco, CA Joined February 2024

Tweets

16K
Followers

2K
Following

599
Likes

18K

Tom Boyle @tomlikestocode

3 weeks ago

@nixcraft Next thing you’ll tell me is netstat isn’t around anymore either…

4 1 102 3K 0

Lisan al Gaib @scaling01

4 weeks ago

GPT-5 with high reasoning effort on SimpleBench

230 229 2K 10.5M 504

Download Image

BREAKING: Apple just proved AI "reasoning" models like Claude, DeepSeek-R1, and o3-mini don't actually reason at all. They just memorize patterns really well. Here's what Apple discovered: (hint: we're not as close to AGI as the hype suggests)

3K 10K 65K 14.1M 42K

Download Image

Guillermo Rauch @rauchg

3 months ago

Build a product so good, your users do the marketing

210 503 4K 244K 522

Tom Boyle @tomlikestocode

3 months ago

Big thanks to @awscloud for an inspiring day.

Swift Compute @swiftcompute

3 months ago

Big thanks to @awscloud for an inspiring day.

0 0 2 320 0

Download Video

0 0 4 257 0

Tom Boyle @tomlikestocode

3 months ago

future dev goldmine

Adina Yakup @AdinaYakup

3 months ago

future dev goldmine

3 19 102 5K 30

0 0 7 273 0

Paul Graham @paulg

4 months ago

I like this image by @lehoho248.

95 649 7K 593K 2K

Download Image

Paul Graham @paulg

5 months ago

Especially when you're young, beware of dismissing some things you're curious about as unimportant. You risk pruning off outliers if you do that. A lot of great discoveries have come from investigating things everyone else thought were unimportant.

89 365 4K 166K 656

Logan Kilpatrick @OfficialLoganK

5 months ago

it’s very easy to gloss over just how hard it is to land models at the frontier that you can also reasonably afford to pay for at scale 2.5 Pro is really something special

176 125 3K 157K 295

Download Image

Guillermo Rauch @rauchg

5 months ago

Same foundation whether you have 1 user or 1B users 📂 app ├ 📄 layout.tsx └ 📄 page.tsx 📂 styles └📄 globals.css 📄 next.config.ts 📄 tsconfig.json 📄 package.json

111 184 5K 377K 2K

NVIDIA @nvidia

5 months ago

Announced at #GoogleCloudNext: NVIDIA and @googlecloud are bringing #agenticAI on-prem with support for Google Gemini models—powered by NVIDIA Blackwell and Confidential Computing for top-tier performance and security. nvda.ws/4loZnhh Hear more from our CEO Jensen Huang.

46 178 727 77K 76

Download Video

Pietro Schirano @skirano

5 months ago

Yeah RAG is dead.

294 115 2K 741K 858

Awni Hannun @awnihannun

5 months ago

First results are in. Llama 4 Maverick 17B active / 400B total is blazing fast with MLX on an M3 Ultra. Here is the 4-bit model generating 1100 tokens at 50 tok/sec:

43 131 1K 149K 426

Download Video

Meta for Developers @MetaforDevs

5 months ago

We’ve updated the official Llama repo on GitHub with new inference code, reference implementations and more for working with Llama 4 ⬇️ llama-models: bit.ly/4hZW0KL llama-stack: bit.ly/4j0qsGb llama-cookbook: bit.ly/4coeOlS

19 225 1K 71K 199

Download Image

Tom Boyle @tomlikestocode

5 months ago

If you’re spending more on inference than on salaries, let’s talk.

0 0 3 79 0

Tom Boyle @tomlikestocode

5 months ago

You don’t need to switch to Swift Compute overnight. Start with 1 job. Feel the difference.

0 0 3 77 0

Andrej Karpathy @karpathy

5 months ago

Let’s take AI predictions from blog posts, podcasts and tweets and move them to betting markets, our state of the art in truth. My struggle has been coming up with good, concrete, resolvable predicates. Ideally, predicates related to industry metrics and macroeconomics. Eg…