Changkun Ou

@MaiYangAI No. I will use Claude or Codex

if price wasn’t a factor, most devs would be on Cursor. Do you agree?

I wrote an essay tracing this line of thought, around AI system design and stability, self-reference, Gödel, entropy, and the conditions under which a system might begin to form a model of itself. Sharing as a thought experiment at the intersection of AI, cognition, philosophy.

Discussions about AI often start with models, compute, and capabilities, but if we follow the questions far enough, they quickly turn into something deeper: what is a "self," how does consciousness arise, and why do some systems begin to recognize themselves at all?

At the Boundary of Self-Reference open.substack.com/pub/changkun/p…

@MaiYangAI 软件工程师的头衔可能会消失，知道应该及为什么要做什么的技能不会

@ouchangkun 应该是什么意思

我真的是服了，相同的一句话，用相同的谷歌翻译，但是结果意思完全不同。你有遇到过吗？

Our friends at @SixtTech has built and open-sourced an MCP Server for @tensorlake to enable enterprise wide Agentic Document parsing. The MCP server exposes the following tools to Agents - Parse documents in 30+ formats to markdown, extract structured data, classify pages in long documents. Their agents use the MCP server autonomously understand knowledge in documents, which then also flow into chat based interfaces and databases for future lookups. Shout out to @ouchangkun for open sourcing the project! He is leading the work of building cutting edge agentic automation workflows for products millions of people use every day. Code - github.com/Sixt/tensorlak…

We should all write papers like this: arxiv.org/pdf/1804.02767

@ilyasut Too much steering suppresses surface symptoms while leaving deeper issues amplified. Instead, openly and transparently presenting both sides well reduces alignment faking. The whole alignment problem look like cultivating a stable attractor, an interpretable equilibrium.

Important work

Final day of #Mum2024 in Stockholm (I am already back in Munich teaching) but wanted to share one more presentations by, Changkun Ou, Felix Putze, and Sven Mayer on detecting internal and external attention with EEG. @francescochios3 @svenmayerhci @ouchangkun @mimuc

@zigo_101 are we arguing about consistency of the panicking behavior in range or arguing about the panicked example is not a suitable example for this consistency discussion? why?

@ouchangkun Because for range nilArrayPointer {} for i := range nilArrayPointer {} for i, v := range nilZeroLengthArrayPointer {} all don't panic. So for i, v := range nilNonZeroLengthArrayPointer {} is a bad example.

In #Golang: * range over a nil slice is a no-op. * range over a nil map is a no-op. * range over a nil array pointer is a no-op. * range over a nil (1.23+) iterator panics.

@zigo_101 because?

@ouchangkun I don't agree with Russ' view and I think his example is bad.

2023 achievements: - Defended my PhD - Most frequent and longest travel records in my life - Starting to understand absurdism and trying to become existentialism

Paper presentation by @francescochios3 on "Short-Form Videos Degrade Our Capacity to Retain Intentions: Effect of Context Switching On Prospective Memory."@LukeHaliburton, @ouchangkun, Andreas Butz, @albrechtschmidt #CHI2023 #GermanPreCHI @GermanHCI arxiv.org/abs/2302.03714

Excited to share the preprint of our new #CHI2023 paper coauthored with @LukeHaliburton , @ouchangkun , Andreas Butz and @albrechtschmidt. Have a look at the teaser video and here at the paper➡️ rb.gy/o8enkm #DigitalWellbeing #SocialMedia #TikTok #Interruptions

@googollee 😄原来大家都认识

一个认识的人，滑雪触发雪崩被埋了，没救回来……

Mathematics 😍

I am very much surprised and impressed by the learned experiences and practices from a hundred-year-old company, where engineers try hard to get rid of five decades old Cobol programs without breaking anything🤯

@oulasvirta Like rationality, rigor remains loosely defined, it depends on the assumptions we have, it depends on the problem scope we focus, it depends on the amount of observation we collected, it depends on the methodologies that are socially accepted, many dependencies…

\section{Rigor}