HN user
CTO @ Xemantic | Metacognitive AI | Harness Engineering | Humanistic & Artistic Computation | co-founder of prachtsaal.berlin
I am shocked how much my experience is different from yours. I wrote Claudine, my own version of Claude Code, almost 2 years ago. This experience gave me the understanding of how the technology works. Since then I've produced maybe 300k lines of open source code, and all of it meaningful to the bones. What kind of projects are you working on, maybe it's the specificity of your domain?
Quite aligned with my own experience from harness engineering and winning AI4Science hackathon. During the hackathon I was working as a human optimizer, moving the feedback from test harness running on Claude Code, back to my local Claude Code for analysis-hypothesis-proposal cycle. And in this moment I realized that 2 Claudes talking to each other could actually scale much better.
I've been teaching harness engineering for a while:
How do you handle possible prompt injection in emails?
https://github.com/xemantic/markanywhere
Incremental Markdown parser that emits streams of semantic events, plus tools to manipulate them - designed for real-time rendering of streamed LLM output.
Regarding applications/industries - after software development as an industry I see scientific research benefiting the most from agentic AI and autonomous machine reasoning. Another domain which is still not saturated would be something I call "personal guardian angel", which usually extends to whole families. Quite often people use OpenClaw for these use cases. Personally I build my own harnesses. One instance is set up as a CEO of my organization, taking care of operations, and another one, completely separate, operating over private knowledge about my family matters, schedules, medical history, etc. I predict context graph and metacognitive use cases to get rapid adoption this year.
- I've built 2 harnesses, one called Claudine - an older sister of Claude Code which now I use for teaching harness engineering, and Golem XIV, with context/knowledge graph management and self-directed metacognitive research.
- In practice rather Anthropic models if the quality of the metacognitive reasoning process and self-improvement loops are considered.
More than a year ago I built my own coding agent called Claudine. I also made agentic anthropic-sdk-kotlin and few other AI libraries for the ecosystem. Maybe this low-level experience allows me to use these tools to deliver in 2 days what would have taken 2 months before.
My advice - embrace TDD. Work with AI on tests, not implementation - your implementation is disposable, to be regenerated, tests fully specify your system through contracts. This is more tricky for UI than for logic. Embracing architectures allowing to test view model in separation might help. I general anything reducing cognitive load during inference time is worth doing.
Suboptimal choice. According to AutoCodeBench, for equivalent problem complexity, LLMs generate correct Kotlin code ~70% of the time versus ~40% for Python, and Go scores lower than Python. Kotlin can be executed as a script while providing super fast compilation phase next to evaluation phase, which is further reducing a chance of mistakes. I don't use tools anymore. I just let my LLMs output Kotlin script directly together with DSLs tailored to the problem space, reducing cognitive load for the machine. It works like a charm as a Claude Code replacement, not only coding autonomously in any language, but directly scripting DB data science, Playwright, etc., while reducing context window bloat.
1. LLMs excell at extracting facts from the context. Storing them as a subject-predicate-object relationships is "natural" for graph databases. Doing it right, so that this knowledge can be utilized more efficiently than any RAG, requires sophisticated context engineering, for example to avoid duplicates and keep consistent vocabulary for relationships, but it is totally achievable and the quality of automatically extracted knowledge can be almost spotless, especially if an LLM can also decide on generating parallel embeddings as a semantic search entry point for graph traversal.
2. Writing cypher queries is a job I would never like to have as a human. But LLMs love it, so that an agent can do an ad hoc data science for every single problem. Especially while being aware which criteria were used for graph construction. It is worth ditching things like MCP in favor of tool graph-like solutions. For this purpose I developed my own DSL which only LLM speaks in internally. The effects are mind-blowing.
I would support every political and social movement progressing us on the spectrum from patriarchy to matriarchy. In particular I would put pressure on the legal system of countries where women are still not allowed to inherit land and property.
The next biggest problems to tackle:
- the way we are producing proteins
- the way we are producing energy
- adoption of cognitive AI in scientific research
Personally I use it for implementing complete new features using test driven development, or fix bugs in existing code base.
Outside of software development I use it for managing all the legal documents of our cooperative (and it's Germany, and I don't speak German well ;) )
It is exactly the opposite. Participants collect all the pieces of knowledge to build an agent like my Claudine at the end of the workshop:
I added this functionality already some time ago in my Claudine agent:
https://github.com/xemantic/claudine/
It costed roughly 30 lines of code: https://github.com/xemantic/claudine/blob/main/src/commonMai...
I am the author of Claudine, who presented the project during this meetup, and I am happy that it is inspiring others. Claudine is intended for educational purposes, so that anyone can easily build a powerful (e.g. Unix-omnipotent) autonomous agent. It is possible thanks to my work on:
https://github.com/xemantic/anthropic-sdk-kotlin
Much more complex custom agents can be built thanks to this library, simplifying the task as much as I could design it.
Please also take a look at my omnipotent Claudine. It is good at self-modifying to develop new tools:
Is there anything special required for brainstorming except for system prompt? You can try providing new system prompt to my Claudine:
I am a shader developer, often struggling with the incompatibility of underlying rendering tech (OpenGL, Vulkan, Metal, etc.). I built my own live-coding visual synthesizer using Kotlin script and OpenGL. If I had some time, I would port it to Google filament, so it is truly multiplatform.
I am happy with JReleaser, I just started migrating all my libraries from publishing through io.github.gradle-nexus.publish-plugin to publishing with jreleaser, which became necessary after migrating my sonatype account to the new maven central scheme.
I also configured the announcement feature, so now I can share each release on my discord server and hopefully soon on LinkedIn.
The documentation of JReleaser is quite comprehensive, however does not fully cover "howto" steps regarding auth for each provider. Which in my case translated to initial cycle of try-and-error with my GitHub actions.
Location: Berlin
Remote: Yes
Willing to relocate: No
Technologies: Agentic AI, LLMs, Kotlin/Multiplatform, Java, GLSL/Shaders, Graphics/GPU programming, JavaScript, Spring Boot, DevOps, AWS, GCP, BigQuery, PostreSQL, Oracle, ETL, Terraform, software/cloud architecture
Résumé/CV: https://www.linkedin.com/in/kpogoda/
Email: in CV
Recently I specialize in teaching how to build AI agents and implementing them in Kotlin multiplatform environment according to my open source SDKs, therefore I can teach, and consult introduction of AI agents, in your organization.
Experienced coder, software architect, creative technologist, computational artist, community builder, founder of Xemantic, a collective of applied philosophy, and Prachtsaal, a non-profit art cooperative from Berlin.
Website: https://xemantic.com
SEEKING WORK | Berlin | Remote
Technologies: Agentic AI, LLMs, Kotlin/Multiplatform, Java, GLSL/Shaders, Graphics/GPU programming, JavaScript, Spring Boot, DevOps, AWS, GCP, BigQuery, PostreSQL, Oracle, ETL, Terraform, software/cloud architecture
Résumé/CV: https://www.linkedin.com/in/kpogoda/
Website: https://xemantic.com
I just won AI Hack Berlin hackathon with the AI agent completely controlling my computer, while programming own tools, and administering Unix system, according to the input in the form of a sketch made by the user.
Recently I specialize in teaching how to build AI agents and implementing them in Kotlin multiplatform environment according to my open source SDKs, therefore I can teach, and consult introduction of AI agents, in your organization.
Experienced coder, software architect, creative technologist, computational artist, community builder, founder of Xemantic, a collective of applied philosophy, and Prachtsaal, a non-profit art cooperative from Berlin.
The point is that we don't have to adhere to "computer use" API at all, and in some situations it might be even counterproductive, while the "bash tool" system prompt comes with implied assumption that we are running inside docker container. Check the system prompt of my Claudine agent. I wonder if this kind of use will be ever constrained by Anthropic? To be honest I am quite surprised that I could do it before.
SEEKING WORK | Berlin/Worldwide | Remote
Technologies: Kotlin/Multiplatform, Java, GLSL/Shaders, Graphics/GPU programming, JavaScript, Spring Boot, DevOps, AWS, GCP, BigQuery, PostreSQL, Oracle, ETL, software/cloud architecture, Anthropic API
Résumé/CV: https://www.linkedin.com/in/kpogoda/
Email: in CV
Location: Berlin
Remote: OK
Willing to relocate: No
Technologies: Kotlin/Multiplatform, Java, GLSL/Shaders, Graphics/GPU programming, JavaScript, Spring Boot, DevOps, AWS, GCP, BigQuery, PostreSQL, Oracle, ETL, software/cloud architecture, Anthropic API
Résumé/CV: https://www.linkedin.com/in/kpogoda/
Email: in CV