Introducing 'ta', a The Archive-compatible Zettelkasten exploration tool for coding agents
ta is a The Archive-like search command-line tool for your Zettelkasten.
Install from https://codeberg.org/ctietze/ta by downloading the tar compressed binary, or the installer .pkg file -- your choice!
(GitHub Mirror at our organization https://github.com/Zettelkasten-Method/ta )
Its intended use is to help with scripting, but also with coding agents (Claude, Codex, Gemini, Grok, Cursor, AMP, ...) to browse your notes.
Command-Line Interface
The command-line tool is simple enough and self-documenting for agents to fiddle with it:
$ ta
OVERVIEW: Zettelkasten retrieval CLI for coding agents.
EXAMPLES:
Find notes by tag, expand 1 hop along wiki-links:
ta search --tag learning --depth 1
AND-combine predicates (tag + phrase + whole word):
ta search --tag thinking --phrase "second-order" --word inversion
Positional phrase shortcut:
ta search "mental models" --depth 2
Tag-only convenience:
ta tag thinking --depth 2
Print one or more notes (frontmatter + raw markdown):
ta show "202503091430 Mental Models.md"
Archive path resolution (highest precedence first):
1. --archive PATH — flag on a subcommand (search, tag, or show)
2. TA_DIR — environment variable
3. archive: PATH — key in ~/.config/ta/config.yaml
USAGE: ta <subcommand>
OPTIONS:
--version Show the version.
-h, --help Show help information.
SUBCOMMANDS:
search Search the Zettelkasten with AND-combined predicates.
tag Search for notes carrying a specific hashtag (convenience for 'search --tag').
show Print the full body of one or more notes.
See 'ta help <subcommand>' for detailed help.
Agent Skills
The real value for users of The Archive is in agent skills. You already have the app, why deal with an inferior experience
But agents don't so they'll benefit.
The repository contains copy-paste-able agent skills in the canonical format of https://agentskills.io/
Find the list here for your convenience:
https://codeberg.org/ctietze/ta/src/branch/main/docs/skills
LLMs and coding agents by extension are good for coming up with similar word searches, and the skills are meant to trigger this capability; see the deep research example below.
Fiddle with the skills, or write your own ones; there's no reason to be precious about these Markdown files. I'm experimenting with the ones I have for a while and report back and invite you to do so, too!
Author at Zettelkasten.de • https://christiantietze.de/
Howdy, Stranger!
Comments
Deep research examples
Some technical connections
Here's a test run dumped from Claude for
deep research my zk for RAG and vector db'sbefore I renamed the tool fromzk-llmtota, hence the weird names in the bash commands:❯ deep research my zk for RAG and vector db's ⏺ Skill(zk-llm-deep-research) ⎿ Successfully loaded skill ⏺ I'll start with a broad search to find seed notes on RAG and vector databases. ⏺ Bash(swift run zk-llm search "RAG" --depth 1 2>&1 | tail -80) ⎿ [0/1] Planning build Building for debugging... [0/3] Write swift-version--6CDFB4424A1E79B2.txt … +26 lines (ctrl+o to expand) ⏺ Bash(swift run zk-llm search "vector" --depth 1 2>&1 | tail -80) ⎿ snippet: "as a result. (stdin to file) Mermaid, graphviz. LaTeX for vector renditions of formulae. There could be more use cas" tags: [] … +58 lines (ctrl+o to expand) ⏺ Bash(swift run zk-llm search "embedding" --depth 1 2>&1 | tail -80) ⎿ snippet: "Look at Backlinks[[202503100826]] and especially structure s embedding the note you change to make sure that the link co" tags: [zettelkasten, api] … +60 lines (ctrl+o to expand) ⏺ Good — I've found the hub notes. Let me read the core ones and look for adjacent concepts like hybrid search. ⏺ Bash(swift run zk-llm show "202603280645 Zettelkasten and RAG.md" "202603280646 RAG Retrieval-Augmented Generation.md" "202603280655 RRF Reciprocal rank fusion.md" 2…) ⎿ Building for debugging... [0/3] Write swift-version--6CDFB4424A1E79B2.txt Build of product 'zk-llm' complete! (0.13s) … +90 lines (ctrl+o to expand) ⏺ Bash(swift run zk-llm tag rag --depth 2 2>&1 | tail -60) ⎿ Building for debugging... [0/3] Write swift-version--6CDFB4424A1E79B2.txt Build of product 'zk-llm' complete! (0.12s) … +26 lines (ctrl+o to expand) ⏺ Let me quickly check the BM25 and tf-idf notes since they're referenced often. ⏺ Bash(swift run zk-llm show "202412071858 Search ranking with tf-idf term frequency -- inverse document frequency.md" "202412071908 BM25 search ranking.md" 2>&1 | ta…) ⎿ Building for debugging... [0/3] Write swift-version--6CDFB4424A1E79B2.txt Build of product 'zk-llm' complete! (0.12s) … +34 lines (ctrl+o to expand) ⏺ I have enough to synthesize. No standalone "vector database" note — your thinking treats vectors as one retrieval mechanism among several. Your thinking on RAG and vector databases The cluster is small and intentional — five notes, written mostly in late 2024 and March 2026, that frame RAG as an architectural pattern and vector search as one pluggable retrieval option inside it rather than the defining feature. Three themes: 1. RAG as a three-stage architecture, with retrieval as the pluggable middle. - 202603280646 RAG Retrieval-Augmented Generation.md — defines the indexing → retrieval → generation pipeline, and explicitly separates the mechanism from the architecture. Vector embeddings are the common choice, but lexical (tf-idf, BM25, SQLite FTS5) "can serve as the retrieval backend." - You note the vector trade-off directly: higher compute, reduced debuggability — in exchange for cross-lingual / semantic hits ("knowledge cycle" matching "Wissenszyklus"). 2. Hybrid search, fused by RRF, is the production answer. - 202603280655 RRF Reciprocal rank fusion.md — you chose RRF because it sidesteps the score-normalization problem (BM25 unbounded, cosine 0–1) by ranking-only fusion. The k=60 dampening makes it so a document has to appear across methods to win. This is a quietly strong claim: you don't think vector search alone is enough. - 202412071908 BM25 search ranking.md and 202412071858 … tf-idf.md — the lexical half of hybrid. You keep the older Zac Romero (EmacsConf 2024) lineage alive here, which is why your RAG notes reach for BM25 as a first-class peer to embeddings. 3. The Zettelkasten/RAG mapping — and where each has blind spots. - 202603280645 Zettelkasten and RAG.md is the synthesis hub. You draw a table aligning atomicity ↔ chunking, tags ↔ lexical index, links ↔ graph retrieval, structure notes ↔ parent-document retrievers, synthesis ↔ generation, GIGO ↔ garbage retrieval. - Your sharpest claim lives here: hand-curated structure notes encode intentionality that embedding similarity cannot replicate. The inverse is equally honest — vector search finds unlinked cross-cluster connections a Zettelkasten has no edge for. You even propose the experiment: embed everything, query with vector search, diff against structure-note traversal. The delta is the research question. The thread connecting all three: you treat vector similarity as a complement, not a replacement — for lexical ranking (via RRF), and for human-authored links (via the ZK↔RAG mapping). The consistent move is skepticism toward any single retrieval signal. Gaps worth noting: no standalone note on vector databases as infrastructure (FAISS, pgvector, HNSW, chunking strategy, embedding model choice). If you want one, 202603280646 is the natural parent to link from. ✻ Crunched for 1m 7sHigher-level stuff
Another example for
deep research my zettelkasten for connections of software architectute and Zettelkasten patterns❯ deep research my zettelkasten for connections of software architectute and Zettelkasten patterns ⏺ Skill(ta-deep-research) ⎿ Successfully loaded skill ⏺ Bash(ta --help 2>&1 | head -50) ⎿ OVERVIEW: Zettelkasten retrieval CLI for coding agents. EXAMPLES: … +25 lines (ctrl+o to expand) ⏺ Bash(ta search "software architecture" --depth 1 2>&1 | head -100) ⎿ - ref: "201606161929 ReSwift doesn't impose architectural constraints. md" title: "ReSwift doesn't impose architectural constraints" … +45 lines (ctrl+o to expand) ⏺ Bash(ta search --phrase "Zettelkasten" --phrase "software" 2>&1 | head -80) ⎿ - ref: "201009241818 Introducing Notational Velocity with MultiMarkdow n support.md" title: "Introducing Notational Velocity with MultiMarkdown support" … +108 lines (ctrl+o to expand) ⏺ Bash(ta search --phrase "pattern language" --depth 1 2>&1 | head -80) ⎿ - ref: "202501301159 § Pattern Language for Zettelkasten.md" title: "§ Pattern Language for Zettelkasten" snippet: "https://notes.andymatuschak.org/z79rMNNuLrUj6R8q1CXSpDm … +120 lines (ctrl+o to expand) ⏺ Bash(ta show "202501301159 § Pattern Language for Zettelkasten.md" "202209211603 § Zettelkasten for programmers.md" "202210081852 Complexity is required in code base…) ⎿ --- ref: "202501301159 § Pattern Language for Zettelkasten.md" title: "§ Pattern Language for Zettelkasten" … +391 lines (ctrl+o to expand) ⏺ Bash(ta show "202501301159 § Pattern Language for Zettelkasten.md" 2>&1 | tail -250) ⎿ - Solution: Write your notes so that they can be composed with other n otes into new texts, making each note a movable text fragment. - Movable type made it possible to assemble pages from smaller pie … +250 lines (ctrl+o to expand) ⏺ Bash(ta show "202209211603 § Zettelkasten for programmers.md" "202210081852 Complexity is required in code bases and Zettelkasten alike.md" 2>&1) ⎿ --- ref: "202209211603 § Zettelkasten for programmers.md" title: "§ Zettelkasten for programmers" … +93 lines (ctrl+o to expand) ⏺ Bash(ta show "202404061208 Avoid implicit themes in your Zettelkasten.md" "202308011922 Locality of change in software.md" "202406132000 Structure vs Process in Zett…) ⎿ --- ref: "202404061208 Avoid implicit themes in your Zettelkasten.md" title: "Avoid implicit themes in your Zettelkasten" … +116 lines (ctrl+o to expand) ⏺ Bash(ta show "202210171300 Zettelkasten as API needs entry points.md" "202111291403 Gall's Law working systems always start simple.md" "202403141330 Zettelkasten tra…) ⎿ --- ref: "202210171300 Zettelkasten as API needs entry points.md" title: "Zettelkasten as API needs entry points" … +51 lines (ctrl+o to expand) ⏺ Software architecture ↔ Zettelkasten in your archive Your thinking on this maps cleanly to four themes, all orbiting one big synthesis note. 1. Christopher Alexander is the explicit bridge. The centerpiece is a deliberately-constructed pattern language: - 202501301159 § Pattern Language for Zettelkasten.md — your largest §-note on the topic. Quotes Alexander 1977, organizes Zettelkasten patterns largest-to-smallest (Zettelkasten → Departments → Structures → Notes → Content → Language → Technology), and lifts pattern names straight from software: "Single Responsibility, Separated Concerns", "Boy Scout Rule", "Immutable Storage", "Rule of Three", "Algorithmic Solutions", "Signal vs Noise Ratio". - 202503090539 Cockburns medical pattern language metaphor components.md — Cockburn's Patterns of Project Management as another import of Alexander's grammar. 2. Software antipatterns port to Zettelkasten antipatterns. - 202404061208 Avoid implicit themes in your Zettelkasten.md — explicitly maps the "incidental data structure" programming antipattern onto your Zettelkasten; proposes "Concept-Free Note-Taking" by analogy to Concept-Free Programming. Heuristic: lack of an address is an indicator the idea isn't expressed directly. - 202308011922 Locality of change in software.md → reused as the structural basis for "Single Responsibility, Separated Concerns" in the pattern-language note. SRP is about change, not essence; atomicity of a Zettel is the knowledge-system analogue. 3. Structure/Process as a unifying principle across both. - 202209232058 Struktur und Prozess von Strukturen und Prozessen in der Programmierung.md — treats Structure×Process as a cross-product applicable to code (static/runtime, module boundaries shifting over coding time) and equally to knowledge work. - 202406132000 Structure vs Process in Zettelkasten for Programmers.md — Atom/Molecule/Organism granularity (Zettel / Structure Note / Project), plus Rule-of-Three extraction before refactoring into a shared theme. - 202111291403 Gall's Law working systems always start simple.md — complexity evolves from a simple working seed, applied to both codebases and Zettelkästen. 4. Zettelkasten as black-box software system. - 202210171300 Zettelkasten as API needs entry points.md — departments are black boxes; structure notes are published API; readers (future-you) interact through curated entry points. - 202209211603 § Zettelkasten for programmers.md — the programmer-facing hub. Ties it together: "Tame complexity like a programmer … Make implicit structures visible with structure notes, similar to extracting a service object." - 202210081852 Complexity is required in code bases and Zettelkasten alike.md — the one-liner version: you don't see benefits of a software architecture before the complexity crosses a threshold — same for a Zettelkasten. The connecting thread visible in the bodies: you treat the Zettelkasten as a long-lived software system. Alexander's pattern grammar (Context/Problem/Forces/Solution/Related) gives the shared formal vocabulary, and programming principles port because both must scale complexity sustainably: SRP ⇢ atomicity, locality of information ⇢ movable text fragments, immutability ⇢ stable addresses, boy-scout rule ⇢ malleable notes, Gall's Law ⇢ start simple and grow, black-box encapsulation ⇢ department entry points. The explicit claim in § Pattern Language for Zettelkasten is that the language itself is the deliverable — a shared vocabulary for people designing tools for thought. If you want to push further, two probes I did not follow: (a) 202308241058 Concept-Free Programming.md paired with the antipattern note, and (b) the "Value Creation in Zettelkasten" / Feynman's Darlings chain from 202406132000, which looked like it connects to a product-engineering frame you haven't fully laid out yet. ✻ Worked for 1m 33sObservations
One thing that sticks out is the limit to
| head -80, limiting the output of searches to 80 lines where each search result takes up a couple of lines can easily miss details.Ideas to improve this:
jsonloneliners, so a limit of 80 would produce 80 results, which is better than ~20 or so you get nowThat's probably Claude Opus 4.7 trying to be clever and save tokens 🤔
Author at Zettelkasten.de • https://christiantietze.de/
Deep Research with GPT-5.4 High Thinking using
piI really really really like the
piagent harness. It's barebones, but you just ask it to create the features you need, and then it happily augments itself. Malleable software! It's a joy to work with.Same prompt as above, including the typo I just noticed now
Since Anthropic killed Claude subscription access outside of their products, I'm using OpenAI here to test everything. The output export is super detailed in comparison to Claude's command line output by default, and GPT is much wordier without a harness making it summarize the findings. See below for the final result.
At the end, GPT-on-pi suggested to create artifacts from the results:
I ordered all three suggestions (attached to this post for you to peek at). The concept map was just another Markdown file, so I then ordered a PNG via graphviz:
Let's just say that @Edmund's graphs look nicer
GPT-on-pi's rather long result
Author at Zettelkasten.de • https://christiantietze.de/
The LLM output here sounds to me like someone who knows @ctietze's thinking.
I removed the link when editing my post by accident, here are the 3 artifacts in text form:
https://gist.github.com/DivineDominion/593e1f6b3412148617368e12f9a6121c
@Andy it does sound somewhat insightful and detailed, and well, it's wordy at least! Was a surprise to read this kind of output, because it's not in my Zettelkasten. And some (obvious in hindsight) connections aren't expressed as cleanly -- like how the similarity of a Zettelkasten and API design fit well with "not everything needs to be public" (in programming) and by conversion not every note needs to be an entry point in the Zettelkasten.
So far it's not a waste of time to use this, but GPT's tendency to be so verbose can quickly turn into getting tired of dealing with the output
Author at Zettelkasten.de • https://christiantietze.de/
Deep research, same prompt, Mistral Vibe using devstral-2
Mistral is a EU provider of LLM services -- apparently heavily subsidized, as they're giving away access to their coding agent for free (!!!!) with some reasonable rate limits.
Mistral replies fast.
If you want to experiment with coding agents, I recommend to set up a Mistral account and then download the
vibeCLI and use that for a taste.Same prompt as above, including the typo:
The reply:
Author at Zettelkasten.de • https://christiantietze.de/
I've sprinkled in some 'deep research my Zettelkasten' sentence fragments in recent programming tasks and the skill detection works well.
Some results genuinely surprise me; the output favors mentioning sources only by ID, which isn't as useful if you can't click them
But traversing the graph is no problem.
So we have:
I haven't tuned the settings, i.e. whether to favor a depth of 1 hop per query, or 3, or 5 or however many. Still experimenting.
For topics where some notes exist, but my thinking is still fragmented, the LLM's overly confident tone shows its flaws. What You See Is All There Is (Kahnemann) on the LLM's side: because there's only a one-sided, two bullet-point overview of a new-to-me topic, it treats this limited perspective as gospel.
So balances need to be added.
Author at Zettelkasten.de • https://christiantietze.de/
Updated recently to make search more similar to The Archive (which means: more loose in a way)
Author at Zettelkasten.de • https://christiantietze.de/
After a few minutes of searching for a .pkg file and then seeing the local build command, I got ta installed. Being an old man, I stumbled with the .config/config.yaml file, but now all is well and running the ta command doesn't invoke an error message. My problem is that no matter which sample command from the help I use, I get an empty [] as the return. Is there a test I can run? Here is the output.
`❯ ./ta search thinking --depth 2
[]`
Will Simpson
My peak cognition is behind me. One day soon, I will read my last book, write my last note, eat my last meal, and kiss my sweetie for the last time.
My Internet Home — My Now Page
@Will thanks for giving it a shot! This is the output I get (knowing that I have notes on the topic
) when I provide a path that exists but doesn't actually contain any text files with that content, e.g.:
But the actual reason could be that you use
.txtfiles and the tool looks for.mdonly. I forgot to push the latest update as a downloadable release. Check out 0.2.0 on the release downloads page: https://github.com/Zettelkasten-Method/ta/releases/tag/0.2.0Author at Zettelkasten.de • https://christiantietze.de/
Still not working. I get a blank set as a return. My zettels use the .md extention. I downloaded and installed "ta-0.2.0-macos-universal.pkg", and I have over 1,200 zettels with "thinking" as a term. Who knew a person could write so much about thinking? Looks of navel-gazing in there, I'm sure.
ta search thinking --depth 2 --archive "~/Dropbox/zettelkasten"Is there something about the title format requirements that might be the issue? My title format is: "Thinking About Thinking - Cognative Distancing 202506252102". The file name is the same with the .md extension.
I tried searching for the term "allostatic," which is not in my zettelkasten yet, and I get a blank set as well. It seems like ta is not searching in my zettelkasten.
Will Simpson
My peak cognition is behind me. One day soon, I will read my last book, write my last note, eat my last meal, and kiss my sweetie for the last time.
My Internet Home — My Now Page
@Will Yes, ID at the end breaks the lookup! Will send an update soon
Author at Zettelkasten.de • https://christiantietze.de/
@Will Check out the latest 0.3.0 please https://codeberg.org/ctietze/ta/releases/tag/0.3.0
Author at Zettelkasten.de • https://christiantietze.de/
This search seems to be the equivalent of the "Concise Bounded Description" (CBD) which was proposed for the Resource Description Framework (RDF). IIRC, it was originally devised by Patrick Durusau. Basically, a CBD means everything that refers to the original target, plus for each of those, everything that refers to them, and so on to a specified depth. There is no reason why the concept has to be limited to RDF.
I have implemented a very rough prototype of a CBD in my system. I'm still experimenting.
Oh joy!! I’m so excited to be part of this team! I ran ta looking for the variable Emerson and found that when I redirected the output to a file, I got a 1.4MB file. When I added the --depth 2 switch, thinking that it would be bigger, I got a 645KB file. It seemed weird until I looked at the output and saw that without a depth parameter, I was getting depth 3.
Feature request: Have the stdout display the number of hits per level.
Now I’m going to look at the ta agent skills. I'm favoring CODEX for the time being. It is going to take this geezer a while to digest the agent instructions in the skills section of ta-deep-research/SKILL.md. Any tips for me?
Will Simpson
My peak cognition is behind me. One day soon, I will read my last book, write my last note, eat my last meal, and kiss my sweetie for the last time.
My Internet Home — My Now Page
First Pass
I got this going! It focused on my zettelkasten, and produced a clear and readable text. I sense I'll be changing the tone the skill's agent is using. This is fascinating.
Codex Prompt:
Do a deep research pass through my notes on "attention horizon". Start broad, follow promising links, then give me themes, tensions, and open questions.
Return
I did a broadtapass onattention horizon`, then followed the strongest links through mindfulness, devotion, algorithmic attention, writing environment, patience, and later notes on the attentional commons.Source base:
Attention Horizon 202012070703,Attention Autonomy 202012151748,A-Attention Is The Beginning Of Devotion 202006230814,Moments Of Attention Just Happen 202107090513,» Algorithms Rule Human Attention 202011201035,» Two Ways Mindfulness Can Improve Life 202011232030,Priority of attention 201902041012,» Has Time Been Well Spent? 202011152024,Shaping my writing environment 202202060916,The Power of Patience 202206132144,A preoccupation is useful for the mind 202601060859,Defending the Attentional Commons 202411300539,Everything is Becoming Television 202510310637.Core Idea
“Attention horizon” seems to mean: an inner orientation instrument that tells you where attention is currently distributed relative to intention.
The aircraft metaphor in the central note is doing real work. It is not just “focus harder.” It is: notice whether your attention is tilted, drifting, too close, too far, too tense, too diffuse, or no longer aligned with what you claim matters. The concept includes foreground/background attention, temporal distance, spatial distance, and moral priority.
Your examples are revealing: kindness should be close on the horizon; meditation off the cushion should be brought closer through practice; writing sits farther out but is actionable. So the horizon is a map of what is psychically near enough to shape behavior.
Themes
1. Attention is identity formation.
Attention Autonomysays what you pay attention to makes you who you are, partly through neuroplastic reinforcement.Algorithms Rule Human Attentionmakes the same point in algorithmic language: your mind has a recommendation system too. You get more of what you click on internally.The implication: attention horizon is not a passive readout. It is self-making infrastructure.
Attention Is The Beginning Of Devotiontreats attention as the beginning of love, care, and nourishment. “The unnoticed can’t be loved” is the strongest compressed claim in this branch. The attention horizon is therefore also an affection horizon: what is close enough to be cared for?This gives the concept an ethical charge. Squandered attention is not merely inefficiency; it is misdirected devotion.
The notes repeatedly move from personal discipline to external design.
Shaping my writing environmentsays control the environment to control writing.Defending the Attentional Commonsscales that up socially: silence is the condition that lets thinking happen.Everything is Becoming Televisionnames the newer threat: infinite passive flow becoming the background of life.So the horizon is not only “inside the head.” It is co-produced by rooms, screens, peers, silence, habits, and commercial systems.
Two Ways Mindfulness Can Improve Lifesays mindfulness helps you ask whether something is worth doing, and also lets unpleasant or unavoidable moments become sites of freedom.Moments Of Attention Just Happenadds a subtler dzogchen move: attention is already moving; the question is what notices the movement.That means calibration is partly intentional and partly receptive. You steer attention, but you also learn to notice its self-arising motion.
The Power of Patienceis a crucial side branch. It argues that real seeing takes time, and that delay can become an active cognitive state. This deepens attention horizon from “what am I focused on?” into “what temporal scale am I capable of inhabiting?”A short attention horizon collapses into immediacy. A cultivated one can hold projects, questions, relationships, and slow transformations.
A preoccupation is useful for the mindmakes writing the way curiosity gathers force. A piece of writing in mind becomes a “windmill of thought”: it catches stray perceptions and makes them coherent. This links strongly withAsk Questions That Concentrate Focus.So one practical answer in your notes is: to change the attention horizon, install a question or writing project that makes the world start presenting relevant material.
Tensions
1. Steering vs letting be.
Some notes say shepherd attention like an adult. Others, especially Huang Po and dzogchen notes, warn against discriminating, pondering, or over-directing the mind. There’s a live tension between attention as disciplined steering and attention as non-grasping awareness.
A synthesis might be: steer the conditions and intentions, then relax the grip inside experience.
Attention And Distraction Are Inversely Relatedcomplicates the binary. It suggests focused and diffuse modes feed each other, and that a “distraction” may move between foreground and background. This challenges a simplistic productivity reading.Maybe a healthy attention horizon is not narrow; it is legible, responsive, and intentionally spacious.
Priority of attentionemphasizes adult responsibility: steer your focus. ButDefending the Attentional Commons,Attention as an economic resource, andEverything is Becoming Televisionshow that attention is aggressively harvested. The tension is between sovereignty and ecology.Your notes seem to reject victimhood, but they also refuse the fantasy that attention is purely individual willpower.
Has Time Been Well Spent?says meaning comes from quality of attention in any moment. ButAttention Horizonexplicitly includes distant plans, traits, habits, and writing. The tension: should attention come home to now, or stretch toward future commitments?Possible synthesis: the future belongs on the horizon, but the only place to adjust the instrument is now.
Open Questions
My read: “attention horizon” is one of your better bridge concepts. It connects mindfulness, writing, anti-distraction, patience, environment design, and moral life without reducing attention to productivity. It says: know what is near, know what is steering you, and keep bringing the worthy things back into view.
`
Will Simpson
My peak cognition is behind me. One day soon, I will read my last book, write my last note, eat my last meal, and kiss my sweetie for the last time.
My Internet Home — My Now Page
To not pollute the output (for tool consumption) this could make sense in a filter, i.e. a piped program so you essentially write
do the search | filter, and the filter would print the count for you. That could be a Python script if you like. To not break the tool's pipe-ability, you could write a script that creates a file with the evaluation results and just passes the original input through.You solved this already but the general recommendation is to just ask the agents to do it for you
Author at Zettelkasten.de • https://christiantietze.de/