Search: twitter.com/karpathy | Heykuki News

HK

Heykuki News

Top New Best Ask Show Jobs

Top New Best Ask Show Jobs

61.

Stanford CS231n deep learning for vision videos taken down due to legal concerns

10 years ago

8 points

62.

Karpathy on Twitter: A fun story of trying to buy one small coffee at Starbucks

5 years ago

8 points

63.

CS231n videos taken down due to legal concerns

10 years ago

8 points

64.

The race for LLM "cognitive core"

a year ago

8 points

65.

print(4 // 0.4) == 9.0

7 years ago

7 points

66.

Autoresearch next step: async collaborative for agents SETI home style

4 months ago

7 points

67.

2025 LLM Year in Review (Andrej Karpathy)

6 months ago

7 points

68.

Power to the people: How LLMs flip the script on technology diffusion

a year ago

7 points

69.

[Andrej Karpathy] Let's reproduce GPT-2, in PyTorch from scratch (nanoGPT)

2 years ago

7 points

70.

Agency > Intelligence

a year ago

6 points

71.

LLMs must one day run in Space

2 years ago

6 points

72.

Most LLMs return the same answer in number guessing game

a year ago

6 points

73.

Deepseek released a new frontier LLM with 11x lesser compute than Meta's largest

2 years ago

6 points

74.

Karpathy: Early Thoughts on Vision Pro

2 years ago

6 points

75.

8 months ago

5 points

76.

Andrej Karpathy on X: "My pleasure to come on Dwarkesh

8 months ago

5 points

77.

Chinchilla (70B) Transformer from DeepMind

4 years ago

5 points

78.

We're missing (at least one) major paradigm for LLM learning

a year ago

5 points

79.

Andrej Karpathy: Agency > Intelligence

a year ago

5 points

80.

NotebookLM is quite powerful and worth playing with

2 years ago

5 points

81.

Let's reproduce GPT-2 (124M)

2 years ago

5 points

82.

Andrej Karpathy on automating software engineering

2 years ago

5 points

83.

Andrej Karpathy: LLMs as the kernel process of a new Operating System

3 years ago

5 points

84.

The open source LLM Cambrian Explosion

3 years ago

5 points

85.

Let's build GPT: from scratch, in code, spelled out

3 years ago

5 points

86.

I don't even code, I prompt. & edit

3 years ago

5 points

87.

The Transformer is a magnificient neural network architecture

4 years ago

TeacherTortoise

5 points

88.

GPT-5 Pro is suited for solving hard problems

10 months ago

4 points

89.

On the "Hallucination Problem"

3 years ago

4 points

90.

Karpathy on Why 'Edutainment' Isn't Real Learning (Seek the Sweat) (2024)

13 days ago

4 points