Obsidian + AI: How to Do It The Right Way  🔗 **Source:** [Obsidian + AI: How to Do It The Right Way](https://www.youtube.com/watch?v=a1FDaoF8Jog) > [!note]+ Document Notes > tags: > - #Obsidian > - #AI > - #PKM > - #KnowledgeManagement > - #Privacy > - #LocalFirst > - #Claude > - #Anthropic > - #Workflows > - #Sensemaking > - #DataOwnership > - #IDIFramework > - #BarbellStrategy > - #Productivity > - #ToolForThought > Author: Nick Milo > Source: YouTube (Linking Your Thinking) > Reference: https://www.youtube.com/watch?v=a1FDaoF8Jog > date created: {{date:YYYY-MM-DD}} {{time:HH:mm}} > > # Title: Obsidian + AI — Doing It the Right Way (Nick Milo) > <!-- deep-read:sections start --> ## Table of Contents: - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [⏱️ 00:00–01:24 — Scope, Tension, and “Sacred Space”](#t-000) - [⏱️ 01:25–02:32 — Why Even Consider AI? Superpowers vs. Voice](#t-085) - [⏱️ 01:45–02:32 — The IDI Framework (Imagine → Discern → Integrate)](#t-105) - [⏱️ 02:33–03:25 — Defensive Barbell: Avoid Overgeneration, Set Privacy Policy](#t-153) - [⏱️ 03:26–04:10 — Offensive Barbell: Reflection, Deep Research, “AI Zone”](#t-206) - [⏱️ 04:11–05:27 — Claude Code + Local Files: Why It Fits Obsidian](#t-251) - [⏱️ 05:36–06:46 — Demo 1: 45-Day Analysis to Separate AI Vault](#t-336) - [⏱️ 06:47–07:33 — Demo 2: Thematic Insights Across Daily Notes](#t-407) - [⏱️ 07:34–08:26 — Demo 3: Keyword Audit (“ideaverse”) & Pattern Table](#t-454) - [⏱️ 08:27–09:16 — Demo 4: Good Friction Between Vaults](#t-507) - [⏱️ 09:14–10:14 — Demo 5: Auto-enrich People Notes with Images](#t-554) - [⏱️ 10:12–10:34 — Workshop and Deeper Build](#t-612) - [⏱️ 10:14–11:33 — Why Not in Core? CEO on Privacy-First Philosophy](#t-614) - [⏱️ 11:33–12:24 — Timeless Tools: Links, Paper, and Simple Mechanics](#t-693) - [⏱️ 12:24–13:34 — Sensemaking Archetypes & Using AI Intentionally](#t-744) - [🗒️ Notes (Detailed, Timestamped)](#-notes) - [📖 References](#-references) - [🗣️ Paraphrase](#-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items (Advocacy/Education/OSPO Alignment + Reusable Prompt)](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🔍 SEO Evaluation](#-seo-evaluation) - [🎯 Conclusion](#-conclusion) ### References: Include a list of sources consulted and cited in the article. - Primary source: YouTube video — https://www.youtube.com/watch?v=a1FDaoF8Jog --- ## 📝 Summary Nick Milo (Linking Your Thinking) explores **how to combine AI with Obsidian** without sacrificing Obsidian’s strengths: privacy, ownership, and a “sacred thinking space.” He proposes a **barbell strategy**—play **defense** (avoid overgeneration, set a clear privacy stance) and play **offense** (use AI for reflection, deep research, and batch operations)—guided by the **IDI framework**: **Imagine → Discern → Integrate**. Practically, he demonstrates **Claude Code** acting on **local Markdown files**, writing to a **separate AI vault** to preserve voice and add healthy friction. Finally, he shares Obsidian leadership’s **privacy-first** stance (no core AI on the roadmap; integrations via plugins are fine if private) and encourages using **AI sparingly and intentionally**, aligned to one’s **sensemaking archetype**. --- ## 📖 Main Content ### <a id="t-000"></a>⏱️ [00:00–01:24 — Scope, Tension, and “Sacred Space”](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=0) - Sets goals: show **how** to use AI with Obsidian “the right way,” defend workflows, and go on offense for **narrow** superpowers. - Notes the conflict: Obsidian is **offline, private, personal**—**AI introduces other voices** into a place intended to preserve **your** voice. ### <a id="t-085"></a>⏱️ [01:25–02:32 — Why Even Consider AI? Superpowers vs. Voice](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=85) - AI excels at **large-scale search** (e.g., scanning hours of audio) and **fast retrieval**—useful if contained. ### <a id="t-105"></a>⏱️ [01:45–02:32 — The IDI Framework (Imagine → Discern → Integrate)](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=105) - **Imagine:** Let AI broaden possibilities (even if some output is wrong). - **Discern:** Keep **editorial judgment**—label untrue but still **useful** directions. - **Integrate:** Connect AI-assisted work back to **your life, values, and projects**. ### <a id="t-153"></a>⏱️ [02:33–03:25 — Defensive Barbell: Avoid Overgeneration, Set Privacy Policy](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=153) - **Don’t overgenerate**: AI text can drown your own writing. - **Privacy spectrum:** - **Local-only** = safest. - **Cloud, no training on your data** = middle. - **Cloud, trains on your data** = riskiest but frontier for experimentation. ### <a id="t-206"></a>⏱️ [03:26–04:10 — Offensive Barbell: Reflection, Deep Research, “AI Zone”](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=206) - Use AI to surface **themes from the last ~30 days**, accelerate **deep research**, and—crucially—**quarantine AI output** in a **dedicated AI vault** to create **good friction**. ### <a id="t-251"></a>⏱️ [04:11–05:27 — Claude Code + Local Files: Why It Fits Obsidian](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=251) - **Claude Code** can **read, analyze, and restructure** local files/folders, trigger **multi-step jobs**, and write back to disk—**perfect for Markdown-based Obsidian**. - **Back up first**; benefit: everything stays **portable Markdown**. ### <a id="t-336"></a>⏱️ [05:36–06:46 — Demo 1: 45-Day Analysis to Separate AI Vault](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=336) - Points Claude Code to vault, asks for **45-day analysis**; result is saved to a **separate AI-only vault**. ### <a id="t-407"></a>⏱️ [06:47–07:33 — Demo 2: Thematic Insights Across Daily Notes](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=407) - Produces ~**1,000-word** synthesis (projects, locations, routines, strategies, breakthroughs) for human review. ### <a id="t-454"></a>⏱️ [07:34–08:26 — Demo 3: Keyword Audit (“ideaverse”) & Pattern Table](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=454) - Counts occurrences (e.g., **“ideaverse” ~238**); compiles patterns into a **table** for inspection and **truth-checking**. ### <a id="t-507"></a>⏱️ [08:27–09:16 — Demo 4: Good Friction Between Vaults](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=507) - Reinforces the **wall** between AI output and your main vault; only **manually curated** material crosses over. ### <a id="t-554"></a>⏱️ [09:14–10:14 — Demo 5: Auto-enrich People Notes with Images](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=554) - Claude Code **fetches images** and fills a dedicated **metadata field** across many People notes—**batch enrichment**. ### <a id="t-612"></a>⏱️ [10:12–10:34 — Workshop and Deeper Build](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=612) - Points to a **workshop** for the full build (AI + Obsidian Bases + advanced KM). ### <a id="t-614"></a>⏱️ [10:14–11:33 — Why Not in Core? CEO on Privacy-First Philosophy](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=614) - Obsidian **roadmap: no built-in AI** right now; **plugins/integrations** are fine **if private** (no training on user data; support for end-to-end encryption). ### <a id="t-693"></a>⏱️ [11:33–12:24 — Timeless Tools: Links, Paper, and Simple Mechanics](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=693) - Emphasizes **links** as the core mechanic; **pen & paper** still powerful; tools are a **stack** to match goals. ### <a id="t-744"></a>⏱️ [12:24–13:34 — Sensemaking Archetypes & Using AI Intentionally](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=744) - Introduces **archetypes** (Inner Guide, Synthesizer, Producer, Creative) and a **quiz**; use AI only where it serves your **dominant archetype**. --- ### 🗒️ Notes - **Philosophy & Tension** - [00:00](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=0) AI can both **threaten** and **augment** Obsidian’s ethos; must preserve **voice**. - [01:02](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=62) Obsidian as **sacred space**: a private “ideaverse” for lifelong learning and values. - **Why AI (Narrowly)** - [01:27](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=87) Superpowers: searching **massive data**, finding **needle-in-haystack** passages. - [01:59](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=119) Accepts that AI can be **wrong** and still **useful** as ideation fuel. - **IDI Framework** - [01:45](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=105) **Imagine:** brainstorm with AI; expand horizons. - [02:09](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=129) **Discern:** editorially **reject/retain** outputs; calibrate trust. - [02:20](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=140) **Integrate:** link results to **roles/projects**; only **manually** promote the best bits. - **Barbell Strategy** - **Defense** - [02:41](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=161) **Avoid overgeneration**—don’t let AI swamp your own words. - [02:51](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=171) **Privacy policy:** choose **local-only**, **cloud non-training**, or **cloud trains on data**. - **Offense** - [03:30](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=210) **30-day reflection** & **deep research** as high-ROI use cases. - [03:47](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=227) Create a separate **AI vault** to enforce **good friction**. - **Claude Code + Obsidian** - [04:11](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=251) Acts on **local files**; can **chain multi-step tasks**. - [05:14](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=314) **Back up** before bulk edits; staying in **Markdown** preserves portability. - **Demos** - [05:36](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=336) **45-day analysis** saved to **AI-only vault**. - [06:47](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=407) ~**1k-word** synthesis: projects, routines, breakthroughs. - [07:58](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=478) **Keyword audit** (e.g., “ideaverse” ~**238** mentions) + **pattern table**. - [08:27](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=507) Reinforces **manual curation** across the vault wall. - [09:14](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=554) **People notes**: auto-fetch **images** → filled **image:** metadata. - **Roadmap & Philosophy** - [10:14](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=614) **No core AI** on Obsidian’s roadmap; plugins/integrations allowed if **private**. - [11:33](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=693) **Links** are the essential mechanic; **paper** remains valuable. - [12:24](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=744) Use AI where it fits **your archetype**; take the **quiz**; more in the next video. --- ## 📖 References - YouTube: **Obsidian + AI: How to Do It The Right Way** — https://www.youtube.com/watch?v=a1FDaoF8Jog --- ## 🗣️ Paraphrase 1) **On preserving voice while leveraging AI** ([01:00–02:32](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=60)) *Original idea:* Obsidian is private, offline, and meant to protect your voice, but AI can still be helpful if you treat its output as speculative and apply discernment. *Paraphrase:* Keep your vault as the home for your authentic thinking, and treat AI as a brainstorming partner whose drafts require your editorial hand before any of it earns a place in your notes. 2) **On the barbell strategy and the AI zone** ([02:33–04:10](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=153)) *Original idea:* Play defense (privacy, avoid overgeneration) and offense (reflection, research). Maintain a separate AI vault so only vetted insights cross over. *Paraphrase:* Set guardrails to prevent AI sprawl, while exploiting its strengths; a dedicated AI workspace creates the healthy friction that keeps your main vault human. --- ## 📰 Newsletter **Title:** Obsidian + AI, Without Losing Your Voice **Blurb:** This week we’re channeling Nick Milo’s rock-solid approach to mixing **AI with Obsidian**. The trick? A **barbell strategy**: defend your voice (avoid AI sprawl, choose a **privacy posture**) and go on offense where it counts (30-day reflections, deep research, batch edits). Using **Claude Code** on **local Markdown** lets you automate safely—and a **separate AI vault** adds just enough friction to keep your ideaverse clean. **Learn web dev corner:** Try scripting metadata updates across Markdown, then review diffs in Git—great practice for **JavaScript/Node** or **Python** file ops. **Community highlights:** PKM folks are sharing vault partitioning tricks; WordPress devs report success using AI for **changelog summarization** and **test data generation**. **New plugins:** Look for **local-first** integrations and **privacy-respecting** AI helpers. 🎥 Watch: https://www.youtube.com/watch?v=a1FDaoF8Jog — Reply with your favorite local-first AI pattern! --- ## ✅ Action Items **A) Reusable analysis prompt tailored to your role (DevRel at GoDaddy, OSS/WordPress educator, aspiring OSPO director):** > *Analyze the attached resource through four lenses: (1) **Open-source sustainability** (funding, governance, local-first privacy), (2) **Contributor onboarding & education** (clear curricula for PHP/JS/HTML/CSS/Git/GitHub; assessment rubrics), (3) **Agency enablement** (playbooks that convert AI efficiencies into client outcomes), and (4) **OSPO readiness** (policies for data privacy, AI usage, and upstream contributions). Produce: a) outcome-aligned learning objectives, b) contributor journey map (first issue → maintainer), c) funding plan (percent-of-revenue or time-tithe), d) metrics (retention, PR lead time, docs coverage, privacy compliance), and e) an inclusivity plan (DEIB + accessibility). End with a 30-60-90 day roadmap and risks/mitigations.* **B) How this video aligns with your objectives:** - **Open-source advocacy:** Promotes **local-first** and **privacy-first** patterns; model policies for **no training on user data** in agency/hosting contexts. - **Contributor engagement & education:** Package the **IDI + barbell** ideas into tutorials: “Quarantined AI Vaults,” “Bulk Metadata Edits Safely,” “Review & Promote Flow via Git.” - **OSPO trajectory:** Draft an **AI Usage Policy** (vault partitioning, provenance, human-in-the-loop), plus **privacy SLAs** for internal tools; codify **governance** for plugin selection and data residency. - **Transparency & inclusivity:** Publish an **annual AI & Privacy impact report** (where AI is used, safeguards, incidents, remediation); include **accessibility** and **DEIB** checkpoints. --- ## 🧪 Hypotheses 1. **Vault partitioning** (AI-only → curated vault) reduces **note pollution** and increases **retrieval quality** vs. single-vault workflows. 2. **Local-first AI** (Claude Code on Markdown) yields **higher user trust** and **adoption** than cloud-training models in PKM contexts. 3. The **IDI framework** improves **knowledge retention** and **idea reuse** vs. unguided prompting. 4. **Batch enrichment** (metadata, keyword audits) increases **link density** and **graph cohesion** without degrading author voice—if guarded by manual promotion. ### 📊 Methodology - **Designs:** A/B tests of **single-vault vs. partitioned** setups; pre/post studies on retrieval accuracy; diary studies on trust. - **Data sources:** Vault diffs, link counts, search success rates, task completion time, user surveys (trust, satisfaction). - **Analysis:** Mixed methods—quant (t-tests/ANOVA), qual (thematic coding of reflections). Track **privacy incidents** and **rollback frequency**. ### ❓ Questions - What **thresholds** (token count, % AI text) should block promotion from AI vault to main vault? - Which **local automations** deliver the best ROI for agencies and educators (e.g., syllabus builders, test-data generators)? - How do **archetypes** correlate with **AI feature adoption** and outcomes? ### 🔮 Future Research - Longitudinal effects of **AI quarantine** on voice/style drift. - **Provenance tooling** inside Obsidian: mark AI-authored fragments for future review. - Assess **privacy posture** impacts on community trust and contribution rates. ### 💡 Implications - Establishes **best practices** for AI in PKM/WordPress education, balancing speed with authorship. - Informs **OSPO policies** on acceptable AI tooling, data handling, and documentation standards. --- ## 🧠 Learning Styles Assessment - **Readability (approx.):** Grade **9–10** (conversational transcript with technical terms). 1) **Visual:** Minimal visuals. *Improve:* diagrams of **vault architecture**, flowcharts for **IDI** and **promotion workflow**. 2) **Aural:** Strong (video). *Improve:* offer **audio chapters**, downloadable **podcast** cut, **TTS** of the summary. 3) **Verbal:** Good explanations, but add **glossary** (archetypes, privacy spectrum). 4) **Kinesthetic:** Add **hands-on labs**: run Claude Code on a sample vault; practice **git diff** review and **promotion**. 5) **Logical:** Present **decision trees** for privacy posture; **checklists** for bulk ops and rollbacks. 6) **Social:** Provide a **forum thread** or **Discord** channel to share vault partitioning templates. 7) **Solitary:** Include **reflection prompts** (“What belongs in my ideaverse?”). 8) **Naturalistic:** Use **ecosystem** metaphor for vault hygiene (compost, pruning, native species = your voice). --- ## 📊 Accessibility Assessment - **Screen reader compatibility:** Ensure **semantic headings** in any written recap; add **alt text** to screenshots of vaults and terminals. - **Keyboard navigation:** Ensure all demos have **keyboard-navigable** steps and captions. - **Contrast & readability:** Use **high-contrast** code blocks; minimum **16px** body text. - **Text resizing:** Make diagrams **responsive**; avoid absolute-positioned text in images. --- ## 🌍 DEIB Assessment - **Cultural relevance:** Include examples from **non-English** PKM communities; avoid tool chauvinism. - **Inclusive language:** Define jargon; use **gender-neutral** terms. - **Neurodiversity support:** Break content into **timestamped chunks**, add **checklists** and **summaries**. - **Representation:** Showcase diverse creators in case studies and screenshots. - **Psychological safety:** Encourage **opt-out** of AI; provide **privacy first** templates. --- ## ✅ Logical Validity **Strengths** 1. **Consistent analogy** ([01:25–02:32](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=85)): AI as a powerful but fallible assistant; supports **discernment** rather than blind trust. - *Why it helps:* Sets accurate expectations; encourages human oversight. - *Enhance:* Add concrete error-rate examples and mitigation steps. 2. **Causal reasoning** ([02:33–04:10](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=153)): Separating AI output → preserves voice. - *Why it helps:* Explains mechanism (good friction) preventing vault pollution. - *Enhance:* Provide before/after samples showing style drift avoided. 3. **Ethical reasoning** ([10:14–11:33](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=614)): Privacy-first stance reduces risk and builds trust. - *Enhance:* Include data retention matrices and audit practices. **Potential weaknesses / Fallacy risks** - **Hasty generalization:** Implying partitioning alone solves voice dilution—depends on **user discipline**. - *Fix:* Add **promotion criteria** and periodic **audits**. - **Appeal to authority:** Referencing CEO stance could be read as conclusive. - *Fix:* Pair with independent **threat models** and user studies. **Tone:** Reflective, pragmatic, privacy-conscious; invites user agency. --- ## 🛡️ DARVO Analysis > **Context:** The video is instructional, not adversarial; still, we examine the rhetoric. 1) **Deny** - **Not present.** No denial of wrongdoing or minimizing others’ claims. Instead, acknowledges AI’s **errors** and **risks** ([01:59–02:20](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=119)). - **Signals:** Terms like “**can be untrue** but **still useful**,” “**privacy spectrum**,” and explicit **backups** emphasize transparency rather than denial. 2) **Attack** - **Not present.** No personal attacks; critiques **practices** (e.g., indiscriminate AI buttons) not **people** ([10:14–11:00](https://www.youtube.com/watch?v=a1FDaoF8Jog&t=614)). - **Signals:** Focus on **principles** (privacy, ownership) rather than adversaries. 3) **Reverse Victim and Offender** - **Not present.** The speaker doesn’t portray themselves (or Obsidian) as a **victim** of users or competitors. - **Signals:** Emphasis on **agency** (“choose your privacy posture,” “good friction”), not blame. **Rhetorical cues countering DARVO:** - **Disclosure of limitations** (backup, errors) counters **deny**. - **Process-focused** critique counters **attack**. - **Shared responsibility** for privacy counters **reverse victimhood**. --- ## 🔍 SEO Evaluation If turning this into a blog recap or documentation page: 1) **Keywords:** Obsidian AI, Obsidian privacy, local AI, Claude Code, PKM workflows, Markdown automation. 2) **Meta description (≤160c):** *Use AI with Obsidian the right way: IDI framework, barbell strategy, local Claude Code, and vault partitioning to protect your voice and privacy.* 3) **Title tag (≤60c):** *Obsidian + AI: A Privacy-First Workflow* 4) **Headers:** H2s for **IDI**, **Barbell**, **Claude Code**, **Vault Architecture**; include primary keywords naturally. 5) **Content length:** Aim **1,500–2,200** words with demos and screenshots. 6) **Links:** - **Internal:** Prior posts on Obsidian, privacy, Git workflows. - **External:** Video link; Anthropic/Claude docs; Obsidian plugin docs. 7) **URL:** `/obsidian-ai-privacy-local-first` 8) **Images:** Descriptive **alt text** (“Diagram: AI vault → human promotion → ideaverse”). 9) **Mobile optimization:** Test code blocks and tables for wrap/reflow. 10) **Engagement:** Add **copy-paste scripts**, downloadable **sample vault**, and **CTAs** for discussion. 11) **Schema:** Use **Article** + **VideoObject** (duration, uploadDate, thumbnail). 12) **Social:** Open Graph/Twitter Cards with a clear **thumbnail** of the vault architecture diagram. --- ## 🎯 Conclusion Milo’s approach squares AI’s strengths with Obsidian’s values: **quarantine AI output**, **review deliberately**, and **only integrate what serves your goals**. The **IDI framework** and **barbell strategy** operationalize this, while **Claude Code** shows the power of **local-first automation** on Markdown. Adopt a **privacy posture**, implement **good friction** with a separate AI vault, and keep your **voice** at the center of your ideaverse. <!-- deep-read:sections end --> [[3. Resources/Readwise/Full Document Contents/Articles/Obsidian + AI How to Do It The Right Way|Full Document Contents]] ## Highlights - In this video, we're bringing together AI and Obsidian. I'm going to talk about how I'm using AI, how I defend my workflows from the negative effects of AI, but also how I go on the offense to use AI in narrowly scoped ways to superpower key parts of my existing workflows. I'll share the main tool that I've picked to combine Obsidian with to make AI work. And finally, we'll look into Obsidian's future. When will AI be added? What has Obsidian CEO actually said about it? And should we really add ([View Highlight](https://read.readwise.io/read/01k4z0h6gcg222q59t7pmbw9wt)) - constantly amazed at how much I've thought and experienced in the past 30 days. And AI is great at reflecting back to me those themes and highlights. Additionally, I want to use AI as the tip of the spear for deep research. But the most important guidance I can give you for having an offensive mindset with AI is to create a dedicated AI zone. Have a clear place for your AI generated content. This way you create a wall of good friction between anything that AI generates and any bit of that that you ([View Highlight](https://read.readwise.io/read/01k4z0h6gpfb1wzyp0g4svfm4b))