Shawn Simister
@narphorium.com
Building AI powered tools to augment human creativity and problem solving in San Francisco. Previously @GitHub Copilot, @Google, 🇨🇦
narphorium.com
narphorium.com
I’m seeing confusion about Claude Skills and why you need them. It sort of feels like Anthropic is building tools that fix problems many people won’t know that they have until next year
www.anthropic.com/engineering/...
www.anthropic.com/engineering/...
November 10, 2025 at 1:26 AM
I’m seeing confusion about Claude Skills and why you need them. It sort of feels like Anthropic is building tools that fix problems many people won’t know that they have until next year
www.anthropic.com/engineering/...
www.anthropic.com/engineering/...
Amp replaced context compaction with a /handoff command to delegate work to agents
ampcode.com/news/handoff
ampcode.com/news/handoff
October 30, 2025 at 2:28 AM
Amp replaced context compaction with a /handoff command to delegate work to agents
ampcode.com/news/handoff
ampcode.com/news/handoff
GitHub is building a mission control that will let you delegate to other agents like Claude Code or Codex
github.blog/news-insight...
github.blog/news-insight...
October 30, 2025 at 2:28 AM
GitHub is building a mission control that will let you delegate to other agents like Claude Code or Codex
github.blog/news-insight...
github.blog/news-insight...
The new Cursor Composer view lets you launch 8 agents at the same time
cursor.com/blog/2-0
cursor.com/blog/2-0
October 30, 2025 at 2:28 AM
The new Cursor Composer view lets you launch 8 agents at the same time
cursor.com/blog/2-0
cursor.com/blog/2-0
These new Windsurf codemaps look a lot like Project Copernicus from @kcieslak.io
githubnext.com/projects/cop...
githubnext.com/projects/cop...
October 13, 2025 at 1:45 AM
These new Windsurf codemaps look a lot like Project Copernicus from @kcieslak.io
githubnext.com/projects/cop...
githubnext.com/projects/cop...
Fun panel discussion on spec-based development with @kiro.dev team.
Especially interested in upcoming features to ground the specs to verify that the work is done correctly
Especially interested in upcoming features to ground the specs to verify that the work is done correctly
October 7, 2025 at 3:45 AM
Fun panel discussion on spec-based development with @kiro.dev team.
Especially interested in upcoming features to ground the specs to verify that the work is done correctly
Especially interested in upcoming features to ground the specs to verify that the work is done correctly
The Trailblazer code is open source so you can try the tool in your favorite VS Code fork: github.com/YanLitao/Tra...
October 3, 2025 at 11:35 PM
The Trailblazer code is open source so you can try the tool in your favorite VS Code fork: github.com/YanLitao/Tra...
Teaching LLMs to Plan: Logical Chain-of-Thought Instruction Tuning for Symbolic Planning
arxiv.org/abs/2509.13351
arxiv.org/abs/2509.13351
September 27, 2025 at 5:07 AM
Teaching LLMs to Plan: Logical Chain-of-Thought Instruction Tuning for Symbolic Planning
arxiv.org/abs/2509.13351
arxiv.org/abs/2509.13351
Love this analogy from the RepoAudit paper. LLMs with huge context windows are like putting all your code on a single PowerPoint slide and then projecting it onto a 1000-foot screen
arxiv.org/abs/2501.18160
arxiv.org/abs/2501.18160
August 30, 2025 at 4:54 PM
Love this analogy from the RepoAudit paper. LLMs with huge context windows are like putting all your code on a single PowerPoint slide and then projecting it onto a 1000-foot screen
arxiv.org/abs/2501.18160
arxiv.org/abs/2501.18160
Neat way to use t-SNE to show how the backlinks improve memory retrieval
August 29, 2025 at 6:14 PM
Neat way to use t-SNE to show how the backlinks improve memory retrieval
I believe that prompt optimizers/compilers are the future. Agent prompts today are like the early days of assembly language. Soon, we'll have compilers and higher level programming constructs built on top of natural language. Eventually no one will write these kinds of low-level prompts by hand
August 13, 2025 at 5:48 PM
I believe that prompt optimizers/compilers are the future. Agent prompts today are like the early days of assembly language. Soon, we'll have compilers and higher level programming constructs built on top of natural language. Eventually no one will write these kinds of low-level prompts by hand
This new prompt optimizer from OpenAI is an exciting development. You can give it a prompt that you've already written and it will suggest improvements to your prompt and explain its reasoning for the changes
platform.openai.com/chat/edit?op...
platform.openai.com/chat/edit?op...
August 13, 2025 at 5:48 PM
This new prompt optimizer from OpenAI is an exciting development. You can give it a prompt that you've already written and it will suggest improvements to your prompt and explain its reasoning for the changes
platform.openai.com/chat/edit?op...
platform.openai.com/chat/edit?op...
Top-down development is where current tools still seem to struggle. As you add more real-world constraints you end up having to do a lot more hand-holding to keep the agent on track
3/4
3/4
August 6, 2025 at 4:59 PM
Top-down development is where current tools still seem to struggle. As you add more real-world constraints you end up having to do a lot more hand-holding to keep the agent on track
3/4
3/4
What I’m calling bottom-up development seems to be where people are seeing the most success. Exploring new features without as many constraints on what the end result should be
2/4
2/4
August 6, 2025 at 4:59 PM
What I’m calling bottom-up development seems to be where people are seeing the most success. Exploring new features without as many constraints on what the end result should be
2/4
2/4
Reality Proxy: Fluid Interactions with Real-World Objects in MR via Abstract Representations
arxiv.org/abs/2507.17248
arxiv.org/abs/2507.17248
August 5, 2025 at 12:48 AM
Reality Proxy: Fluid Interactions with Real-World Objects in MR via Abstract Representations
arxiv.org/abs/2507.17248
arxiv.org/abs/2507.17248
Excited to pore over this new book from Buddy Bindery. Thank you @alexanderobenauer.com for sharing your artistry with the world
July 25, 2025 at 1:51 AM
Excited to pore over this new book from Buddy Bindery. Thank you @alexanderobenauer.com for sharing your artistry with the world
Natural Language Outlines for Code: Literate Programming in the LLM Era
arxiv.org/abs/2408.04820
arxiv.org/abs/2408.04820
July 10, 2025 at 1:26 AM
Natural Language Outlines for Code: Literate Programming in the LLM Era
arxiv.org/abs/2408.04820
arxiv.org/abs/2408.04820
It's a side-by-side eval widget where an LLM tries to highlight the key differences to make the user's job easier
July 3, 2025 at 7:20 PM
It's a side-by-side eval widget where an LLM tries to highlight the key differences to make the user's job easier
Knowledge cycles are another wat to visualize this process tis.so/cycles-of-kn...
July 1, 2025 at 2:11 PM
Knowledge cycles are another wat to visualize this process tis.so/cycles-of-kn...
This is an interesting look at the traceable text design pattern by @hitakam.bsky.social and @andrewhead.bsky.social dl.acm.org/doi/10.1145/...
April 27, 2025 at 1:03 AM
This is an interesting look at the traceable text design pattern by @hitakam.bsky.social and @andrewhead.bsky.social dl.acm.org/doi/10.1145/...
Nubya Garcia at the Chapel. One of the best performances I've experienced in a while
April 18, 2025 at 6:45 AM
Nubya Garcia at the Chapel. One of the best performances I've experienced in a while
I think many people don't use custom cursor rules because they don't know where to begin. I'm experimenting with a set of "metarules" which I can use in chat to generate custom rules for any project
April 16, 2025 at 5:10 AM
I think many people don't use custom cursor rules because they don't know where to begin. I'm experimenting with a set of "metarules" which I can use in chat to generate custom rules for any project