Jake Krajewski

☧ Cognitive Science, A.I. Engineer. Formerly @GoPro Product Design. Exploring the intersection of tech, design, and a.i./deep learning.

[Jake Krajewski @Jakekrajewski]

I’m not saying you’re wrong. I’m saying evolve.

[Jake Krajewski @Jakekrajewski]

@FractalAppsCom I appreciate the candid reply. I think genuine perspectives will be what is truly valuable in the next half-decade. At least when the pendulum swings back from the swarm of AI content. I wish you the best though, fellow human!

[Jake Krajewski @Jakekrajewski]

@FractalAppsCom Second sign, the short statement triplet:

“One map. Thirty seconds to orient. Every major architectural decision readable from the terrain alone.”

[Jake Krajewski @Jakekrajewski]

@FractalAppsCom Quickest way to know AI wrote something

Arbitrary mic drop moments:

“The STS uses exactly five layers. Not ten. Five.”

[Jake Krajewski @Jakekrajewski]

The biggest problem I'm currently facing with engineering agent harnesses is overfitting to test cases.

[Jake Krajewski @Jakekrajewski]

@maddycr3 @HowToAI_ I detest the hyperbole that plagues every peddler on the internet space. It’s akin to “step right up, step right up! You are not gonna believe this!”

[Jake Krajewski @Jakekrajewski]

@gerardsans @pors This is Twitter. I think you forgot.

[Jake Krajewski @Jakekrajewski]

@gerardsans @pors Can you provide a single example? <-- better question.

[Jake Krajewski @Jakekrajewski]

Tweaking attack mechanics with @godotengine

Tales of the Lost Raiders: Curse of the Heartstone coming soon on Steam.

[Jake Krajewski @Jakekrajewski]

@gerardsans @pors A single example would be super supportive of your position. There is nothing in a neuron that cannot be modeled mathematically. A position of ”it’s impossible because I am very smart” seems shaky at best.

[Jake Krajewski @Jakekrajewski]

@gerardsans @pors I wouldn't say this is accurate at all. I can think of several ways neurons can be functionally modeled and employed digitally. One would think you've proven otherwise?

[Jake Krajewski @Jakekrajewski]

@ihtesham2005 using godot to build my first game right now!

[Jake Krajewski @Jakekrajewski]

Building a new low poly game with @godotengine

Does using stylized explosions take away from game experience? Or does it need to have smoke, fire, and damage to be fun?

[Jake Krajewski @Jakekrajewski]

Prompt style guides for image generation are insane.

❤️‍🔥Putting together some storyboards for a new game with Figma + GPT Image 2.

[Jake Krajewski @Jakekrajewski]

@alt_w_v_g @pwang I’m sure everyone is impressed on how you treat your private family matters as LinkedIn victory stories.

[Jake Krajewski @Jakekrajewski]

@gpt_alex @draparente That’s one of the main ways I build skills.

[Jake Krajewski @Jakekrajewski]

@draparente I like your style

[Jake Krajewski @Jakekrajewski]

The Effort display in Claude Code should never disappear. It's not visual noise. The fact that it's combined with "For this session only" makes it seem ephemeral, and the fact that it's not displayed makes it seem like you need to constantly make sure

@jenny_wen @joellewenstein

[Jake Krajewski @Jakekrajewski]

Baysia the Ghost Ray

[Anthropic @AnthropicAI]

New Anthropic research: Natural Language Autoencoders.

Models like Claude talk in words but think in numbers. The numbers—called activations—encode Claude’s thoughts, but not in a language we can read.

Here, we train Claude to translate its activations into human-readable text.

[Jake Krajewski @Jakekrajewski]

/peek let's your parallel agents see what each other is doing without needing to have full context

https://t.co/8OtlGpjUyM

@ClaudeDevs @claudeai @jenny_wen