HNHacker News
TopNewBestAskShowJobs

diwank

2,120 karma · joined September 11, 2011

Dropped out of Columbia (physics, philosophy) for a Thiel Fellowship. Built Julep AI — open-source AI agent framework, 7k stars on GitHub — because I wanted agents that could actually do things. Now building Memory Store (YC X26): persistent memory for AI agents. Turns out the hard problem isn’t making agents smart, it’s making them remember.

Still fascinated by what foundational models and cognitive architectures teach us about our own minds. Building beats theorizing, but sometimes they’re the same thing.

https://memory.store

https://diwank.space

hi@diwank.space

submissionscomments

The last human choice: surrendering agency to AI [video]

youtube.com·2 pts·diwank·
0

Physics of Language Models: The Magic of Canon Layers

papers.ssrn.com·3 pts·diwank·
0

Tap into the "Hemingway effect" to finish what you start

bigthinkmedia.substack.com·27 pts·diwank·
3

Bluefin: Linux Workstation OS

projectbluefin.io·2 pts·diwank·
0

The Accidental Art of Writing Shit That Matters

diwank.space·2 pts·diwank·
0

Show HN: SteadyText: Deterministic LLMs: Same input → same output, every time

steadytext.julep.ai·6 pts·diwank·
0

The Accidental Art of Writing Shit That Matters

diwank.space·1 pts·diwank·
0

An early preview of model capabilities – Generalist [video]

youtube.com·2 pts·diwank·
0

The inaugural Llama Startup Program cohort

ai.meta.com·1 pts·diwank·
0

S-Curves and the Bitter Lesson: The Coming AI Plateau

diwank.space·2 pts·diwank·
0

AGI: Levels of Intelligence

diwank.space·1 pts·diwank·
0

Field Notes from Shipping Real Code with Claude

diwank.space·354 pts·diwank·
99

A Vision for Human-AI Co-Evolution from Agentlets to Personoids to ASI

diwank.space·1 pts·diwank·
0

Learned pain as a leading cause of chronic pain – LessWrong

lesswrong.com·6 pts·diwank·
0

Does AI Progress Have a Speed Limit?–Asterisk

asteriskmag.com·2 pts·diwank·
0

Veritasium: What Everyone Gets Wrong About AI and Learning [video]

youtube.com·2 pts·diwank·
0

Show HN: Open Responses –Self-hosted OpenAI Responses API, works with all models

github.com·1 pts·diwank·
0

Show HN: Open Responses – Self-hosted OpenAI Responses API, works with any model

github.com·23 pts·diwank·
2

The Standard Cosmology Model May Be Breaking

physics.aps.org·2 pts·diwank·
0

What satellite images reveal about Myanmar's quake [video]

bbc.com·49 pts·diwank·
4

Qwen3 Is Dropping Soon

github.com·5 pts·diwank·
0

Laboratory Escapes and "Self-Fulfilling Prophecy" Epidemics [pdf] (2014)

armscontrolcenter.org·2 pts·diwank·
0

Wealth, Shown to Scale

eattherichtextformat.github.io·9 pts·diwank·
4

Prism: Manipulating Concepts in Latent Space

thesephist.com·3 pts·diwank·
0

NoCode: Way to write secure and reliable applications

github.com·2 pts·diwank·
0

Nvidia Fugatto 1 – Foundational Generative Audio Transformer Opus 1

research.nvidia.com·8 pts·diwank·
1

DeepSeek launches o1-preview style reasoning model: R1-Lite-Preview

api-docs.deepseek.com·2 pts·diwank·
0

Google renders the font if you search for its name

google.com·3 pts·diwank·
0

Copilot Actions, new agents, and tools to empower IT teams

microsoft.com·2 pts·diwank·
0

The Surprising Effectiveness of Test-Time Training for Abstract Reasoning

arxiv.org·3 pts·diwank·
0
← PreviousPage 2 of 8Next →