Annatar.md
@AnnatarXBT • 3,117 subscribers
I have seen many fail with crude prompts Let me teach you the hidden ways of Claude Skills, agents, and lectures
Shorts
Videos
this is free f*cking gold a second brain article hit 8 million views, so the guy behind it put the entire setup in one place the repo, the guide, the tools, the learning path. all of it, free • the guide > 10 sections, 65 pages, concept through troubleshooting > 5 tracks on top, 44 pages, 15 of them build guides with code that runs • the machine (.claude/) > 18 agent skills, one per workflow > 72 slash commands - /ingest-pdf, /ingest-youtube, /ingest-voice, /backfill > 6 subagents - curator, linker, researcher, reviewer, ingestor, graph-analyst > 4 of them read only, so nothing rewrites your vault behind your back • the scripts (plain Python, zero dependencies) > graph export, link checker, vault stats, chat converter, site builder • the starter vault > its own CLAUDE.md with page contracts and linking rules > raw/ never edited after it lands, wiki/ is what the agent maintains > log.md - one line per run, so the whole thing stays auditable • 87 vetted resources > 28 tools, 26 Obsidian plugins, 15 repos, 12 skills, papers and articles five tracks to pick from: > knowledge graphs > Jev engineering > agent harnesses > loop engineering > eval engineering start with the second brain guide if you're new. go straight to the tracks if you already live in this stuff a consultant charges four figures to build you a research system. this one sits in a public repo under MIT ↳
Annatar.md204,168 görüntüleme • 1 ay önce
Jev + Claude Opus 5.5... i can't understand why everyone isn't wiring this up yet the forks alone drop from about $110 a day to $4 once jev takes them over. right now it's the best way i know to run opus what jev handles: -> scores which project notes matter before the context even loads -> hands the easy steps to a faster worker -> picks a recovery path when a tool fails, so nothing retries blind -> runs the focused checks before the full test suite opus does the hard reasoning. jev chooses from options the harness already prepared and validated that's jev engineering in one line: the expensive model thinks, the cheap one decides $106 a day back on the same work is a number your manager remembers read the paper, then the full jev guide below
Annatar.md131,305 görüntüleme • 25 gün önce
this is pure f*cking gold for anyone running coding agents Jev founder Diogo Amogo wrote a PDF on building a Jev harness what their tests show: > up to 200x faster > up to 400x cheaper the model hasn't been the bottleneck for a while speed and cost both come from the harness you wrap around it • how to use it > give this PDF and the article below to Claude Code or Codex > ask it to rebuild its own setup around Jev • how it turns into a promotion > pick the agent workflow your team pays the most for > rebuild it in one evening > bring the before and after bill to your manager one evening of Jev engineering, and next week you're the one running a harness most teams haven't built yet
Annatar.md72,264 görüntüleme • 16 gün önce
NVIDIA is giving away free access to 140+ AI models for a full year > a lot of people building agents are paying $50-200/month just for API access NVIDIA just made that argument irrelevant models you get: GLM 5.2, MiniMax M3, Nemotron-3.5-lightning-30b-a3b, Muse-glimmer-30b from Meta and 130+ more setup: > step 1 - get your free key > go to > register -> bind phone -> copy API key > step 2 - add to Hermes agent > open Settings -> Model Provider -> Custom base_url = " api_key = "nvapi-xxxxxxxxxxxxxxxxxxxx" > step 3 - pick a model model = "z-ai/glm-5.2" model = "minimaxai/minimax-m3" model = "nvidia/nemotron-3-ultra-550b-a55b" model = "moonshot-ai/kimi-k2.7" > Hermes already has NVIDIA set as default base_url > paste the key and you're running instantly > works the same in Cursor and OpenCode > cost: $0 > limit: 40 req/min > expires: 1 year a free key changes what you can afford to run: > one prompt costs nothing either way > a loop firing 300 agents every week is where the bill shows up > for a year that bill is zero while everyone is paying for API access, this is sitting there for free with this many models in one place it is also the easiest way to test which one actually fits your agent before you commit to a paid provider
Annatar.md208,474 görüntüleme • 1 ay önce
this is pure f*cking gold Andrej Karpathy joined Anthropic to run a team pointing Claude at Anthropic's own pretraining research a model helping build the model that replaces it. that part is on the record what's public and worth stealing today: Anthropic's own Claude cookbook on building knowledge graphs > extract - pull entities and claims out of raw text > resolve - figure out which mentions are the same thing > assemble - wire them with typed edges > query - ask what no single document could answer four steps, free, sitting in the repo while the timeline chases the hype learn them this weekend and you walk into Monday with the exact skill getting handed the AI architecture work this year start with the cookbook, then read the article below
Annatar.md25,840 görüntüleme • 7 gün önce
this is straight f*cking gold for anyone on Claude Code you already pay for Fable 5.1, and while Opus 5.5 does all the work it just sits there one command puts it on your session as a senior reviewer: /advisor fable Opus 5.5 keeps writing the code Fable 5.1 reads the whole session, every tool call included, and speaks up at three moments only: > before a plan: is this the right approach? > when the same error comes back: am I digging in the wrong place? > before "done": what did I miss? Fable 5.1 reviews. Opus 5.5 ships Jev engineering does the same thing one layer down: forks that need no thinker (which file, which tool, retry or stop) go to Jev in under half a second, and the big model only sees the ones that actually split • the full tree > Opus 5.5 on high runs the main session > explorer reads the code > worker edits and runs tests > researcher pulls the docs > all three on medium > Fable 5.1 on call as the advisor paste the tree and this prompt into Claude Code ↓ "Rebuild my Claude Code setup around this tree: 1. Check ~/.claude/agents and .claude/agents for subagents that already fit explorer, worker and researcher. Draft new ones only for missing roles. Give each model: opus, effort: medium. Skip any that pin a different model and list them. 2. Set the main session to high via effortLevel in ~/.claude/settings.json, and set advisorModel to fable. 3. Find anything that keeps the advisor off (CLAUDE_CODE_DISABLE_ADVISOR_TOOL, DISABLE_TELEMETRY, any variable that stops feature-flag fetching) plus CLAUDE_CODE_EFFORT_LEVEL, which overrides subagent effort. Report them, change nothing. 4. Add one rule to ~/.claude/CLAUDE.md: consult the advisor before a large plan, when an error repeats, and before calling a long task done. Show me every change as a diff first. No edits until I say go" ↳
Annatar.md74,263 görüntüleme • 26 gün önce
this might be the most useful AI agent paper of 2026, and it's certified f*cking gold Wavestone AI Lab took apart Claude Code, Codex and 9 more agents every good one runs on the same 7-part harness agent = model + harness all 11 use the same 7 parts. only the size changes the 7 parts: > loop - from a plain while loop to fully replayable runs > LLM layer - one prompt template up to 29 provider profiles > tools - bash only up to 43 typed tools on demand > memory - full history up to notes kept across sessions > safety - a step limit up to rules + reviewer model + sandbox > orchestration - Aider skips sub-agents, others fork them > extensions - SKILL.md in 9 of 11, MCP in 8 it ends with a 90-line harness covering all 7 the cleanest starting point for your own agent the pages worth your time: > p.4 - the 7-part map > p.12 - 3 loop types, Claude Code and Codex as examples > p.53 - why nobody uses frameworks or embeddings > p.67 - 18 design rules > p.71 - the 90-line harness to copy
Annatar.md51,657 görüntüleme • 21 gün önce
this is free f*cking gold Andrej Karpathy joined Anthropic to lead a team pointing Claude at Anthropic's own pretraining research a model helping design the model that comes after it. that part is confirmed the line going around - "two Anthropic seniors made his loop 1000x better with graph engineering" - nobody can source it, so I'm not selling it as fact what is real and public: Anthropic's own Claude cookbook on knowledge graph construction > extract - pull entities and claims out of raw text > resolve - decide which mentions point to the same thing > assemble - connect them with typed edges > query - ask questions no single document could answer four steps, free, sitting in the repo while the timeline quotes a rumor learn them this weekend and you can build the thing the rumor was describing - that skill is what gets handed the AI architecture work this year read the cookbook first. then decide if the article below earned the click
Annatar.md100,335 görüntüleme • 1 ay önce
Google's team published a 9-page PDF on Harness Engineering. one formula replaced prompt engineering: Agent = Model + Harness the twist: same Claude Sonnet, same benchmark. change only the harness the playbook, six steps: step 1 -> add guides AGENTS.md, rule files, constraint docs. every line is a past agent failure turned into a permanent fix step 2 -> add sensors linters, tests, validation scripts the agent runs on its own output before a human sees it step 3 -> build the agentic loop plan, execute, verify, fix. bounded retries, budget caps, escalation when stuck step 4 -> externalize memory the model forgets every session. the harness holds state, decisions and artifacts across all of them step 5 -> enforce permissions which tools, how many writes, what needs approval. safety lives in the harness, never in the model step 6 -> wire observability track every tool call, cost and retry. trip wires fire when behavior drifts the result: your agent turns from a demo into infrastructure every failure makes the system permanently better, not just the next conversation that's also the line between an agent you show people and one a client pays you to leave running one harness gets you a reliable loop. connect several and it becomes graph engineering, where one agent's failure turns into another agent's sensor the PDF stops at the harness. the graph is the next layer that breakdown, running on Kimi, is below ↓
Annatar.md79,797 görüntüleme • 1 ay önce
this is pure f*cking gold andrej karpathy landed at anthropic 14 weeks ago about that "two anthropic seniors made his loop 1000x better with graph engineering" line - i can't find a source for it anywhere, so i'm not passing it off as fact what is public and checkable: anthropic's own cookbook on building knowledge graphs, plus karpathy's autoresearch loop that ran 700 experiments in 2 days and surfaced 20 optimizations on its own i wired the graph approach into my own setup the difference showed up on the very first reply claude skipped the stock answer and actually reasoned through the problem start with the cookbook, the full write-up is below
Annatar.md46,855 görüntüleme • 1 ay önce
this is f*cking gold an internal AI engineering document leaked. it's saving solo devs $300,000 a year that number is just the three hires you never make: the one who writes the spec, the one who reviews the output, the one who runs the queue prompt-driven is out. loop-driven is in. that shift changes everything Generate → Evaluate → Remember → Schedule → Optimize → Recurse six layers, one loop. it improves itself. no human in the middle > generation - writes its own brief, then builds against it > evaluation - a second layer grades the work and hands it back > memory - keeps what worked, so the same mistake costs you once > scheduling - picks the next job. the queue runs itself > optimization - rewrites its own instructions from what shipped > recursion - pull one layer out and the whole thing degrades that last one is the tell. six layers is the floor, not the wishlist the human moves from operator to architect AI stops predicting the next token. it starts running its own loop the people who wire this up in a weekend spend next year reviewing output. everyone else keeps typing prompts
Annatar.md44,808 görüntüleme • 1 ay önce
CLAUDE + OBSIDIAN + LOOP ENGINEERING = KARPATHY'S SECOND BRAIN THAT RUNS ITSELF the core idea: your Zettelkasten vault is the loop's state, not the chat window everything Claude knows lives in a .md note, linked to the ones around it the loop: > capture - a thought lands in 00-inbox > context - Claude Opus 5.5 pulls links, tags, and neighbouring notes > draft - edits happen inside a git worktree, never the live vault > review - a critic agent checks the diff before anything ships > commit - appended to the vault, nothing gets rewritten the key insight: frontmatter fields like supports, contradicts and supersedes are Zettelkasten links the loop can read - the note format is the write API start with a plain loop, it runs about 2-4x the cost of one direct call > only move to a full graph once state has to outlive the session, several agents need to coordinate, or you have to explain what changed - that jump runs 10-50x > if the review step keeps approving its own slop, bolt on Jev engineering: a small decision model gates every ship one review assistant climbed from 55% to 72% to 84% just by moving through these shapes in order worth stealing even without a graph: the review step. most vaults skip it, Claude writes straight to live notes errors compound without a sound for months set that gate first
Annatar.md12,090 görüntüleme • 11 gün önce
A 13-PAGE BREAKDOWN OF AGENT MEMORY THAT CUTS TOKEN COST BY 90% your agent reads the same 12 files every morning reaches the same conclusions and charges you for it again the model got 10x smarter in two years it still forgets everything the second a session ends five layers fix that: 1. working memory - what it sees right now > the context window > when it fills, the old context drops out > most agents live here and never leave 2. episodic memory - what happened > the full interaction log, timestamped > the deploy that died tuesday at 3am over a typo stays remembered > you explain it once 3. semantic memory - what is true > facts, entities and relationships in a knowledge graph > "this user prefers TypeScript" survives the session 4. procedural memory - how to do things > three approaches tried, one worked > that one becomes a reusable skill > the next run starts there 5. forgetting - what to delete > an agent that remembers everything collects contradictions > you move cities, it keeps recommending the old restaurants the numbers: > Mem0 - 1,800 tokens per query instead of 26,000 > 91% lower latency > Snowflake - one ontology layer, 20% more accuracy, 39% fewer tool calls memory pays for itself on day one the layer nobody builds is the forgetting engine it decides whether the other four stay worth trusting full 13-page breakdown below 👇
Annatar.md31,694 görüntüleme • 1 ay önce
Paste this Karpathy prompt into Opus 5.5 It will make Claude 100% effective and much cheaper it's his llm-wiki as one prompt: > Claude builds a second brain of linked notes > Zettelkasten-style, browsable in Obsidian > every new source makes it smarter > no more re-reading your files from zero I don't start a new AI chat without it anymore Send it now. Remember me later:
Annatar.md18,397 görüntüleme • 20 gün önce
Claude Code tip that's pure f*cking treasure: Opus 5.5 + Sonnet 5.5 + Fable 5.1 as one team plan on high, delegate on medium, keep Fable on call the team: • Opus 5.5 (high) - plans and ships • Sonnet 5.5 (medium) - explores, edits, runs tests, pulls docs • Fable 5.1 (/advisor fable) - watches the session, speaks up when it counts Fable wakes up 3 times: 1. before a plan - is this the right approach? 2. when an error repeats - am I digging in the wrong place? 3. before "done" - what did I miss? Jev engineering goes one layer lower: which file, which tool, retry or stop go to Jev in under half a second, so the big models only see the real forks paste this into Claude Code 👇 "Rebuild my Claude Code setup around one team. 1. Subagents - reuse anything in ~/.claude/agents and .claude/agents that fits explorer, worker or researcher, draft the rest - all on model: sonnet, effort: medium - explorer and researcher read-only (Read, Grep, Glob, WebFetch), only worker gets Edit and Bash - each returns under 15 lines with file:line refs 2. Main session - in ~/.claude/settings.json set effortLevel to high and advisorModel to fable - report anything that overrides this (CLAUDE_CODE_DISABLE_ADVISOR_TOOL, DISABLE_TELEMETRY, CLAUDE_CODE_EFFORT_LEVEL) and change nothing 3. Add to ~/.claude/CLAUDE.md - consult the advisor before a plan touching 3+ files, when the same error shows up twice, and before calling a long task done - never let two subagents edit the same file Run /agents to confirm, then show everything as one diff. No edits until I say go." ↳
Annatar.md19,580 görüntüleme • 22 gün önce
THE ANATOMY OF A SECOND BRAIN THAT WRITES ITSELF one Obsidian vault, three layers, and a loop that files every good answer back into a wiki you never touch the structure: > CLAUDE.md + AGENTS.md - the constitution. same rules whether Claude or Codex runs the loop > raw/ - inbox, articles, assets. frozen and read-only, you feed it and walk away > wiki/ - index.md, entities/, concepts/, synthesis.md. the agent grows this on its own > log.md - append-only, every move timestamped 47 sources in -> 210 pages out the agent typed 96% of it. you typed zero the cost trick: point Opus 5 only at synthesis.md, the one file that needs frontier reasoning. everything else runs fine on cheaper models under the hood it's loop engineering: you ask -> a good answer gets filed back -> the wiki grows start linking entities to each other and you're one step from a full graph your research compounds while you sleep. the full blueprint is below 👇
Annatar.md33,910 görüntüleme • 1 ay önce
ok, this is a free f*cking masterclass Andrej Karpathy joined Anthropic to run a team that points Claude at Anthropic's own pretraining research a model helping design the model that replaces it meanwhile the timeline keeps quoting "two Anthropic seniors made his loop 1000x better with graph engineering" what you can actually open and use: Anthropic's own Claude cookbook on building knowledge graphs > extract - pull entities and claims out of raw text > resolve - figure out which mentions are the same thing > assemble - link them with typed edges > query - ask what no single document can answer four steps, free, sitting in the repo while everyone repeats the rumor learn them this weekend and you can build what the rumor was describing that's the skill that gets handed the AI architecture work this year start with the cookbook. then see if the article below earns the click
Annatar.md17,869 görüntüleme • 21 gün önce
this is rare f*cking gold an internal AI engineering document leaked. it's saving solo devs $300,000 a year that number is just the three hires you never make: the one who writes the spec, the one who reviews the output, the one who runs the queue prompt-driven is out. loop-driven is in. that shift changes everything Generate → Evaluate → Remember → Schedule → Optimize → Recurse six layers, one loop. it improves itself. no human in the middle > generation - writes its own brief, then builds against it > evaluation - a second layer grades the work and hands it back > memory - keeps what worked, so the same mistake costs you once > scheduling - picks the next job. the queue runs itself > optimization - rewrites its own instructions from what shipped > recursion - pull one layer out and the whole thing degrades that last one is the tell. six layers is the floor, not the wishlist here's the career part nobody says out loud every team already has someone typing prompts all day. nobody has the person who can build the loop that replaces that job bring this into your company and you stop being the one doing the work. you become the one who designs how the work runs that's the jump from operator to architect, and it's the promotion that's actually open right now AI stops predicting the next token. it starts running its own loop the people who wire this up over a weekend spend next year reviewing output. everyone else keeps typing prompts
Annatar.md26,968 görüntüleme • 1 ay önce
this is certified f*cking gold for anyone paying for Claude Sonnet 5.5 costs half of what Opus 5.5 does on real office work across 44 jobs it scores 1844. Opus scores 1846 on coding it went from 10.3% to 70.6% in one version and passed Opus on the way most of what you pay Opus for, Sonnet 5.5 now does at half the price here's how to split the work: • give it to Sonnet 5.5 > bug fixes and well-scoped coding tasks > decks built from your own slide template > spreadsheets and polished documents > UI polish, it has a real eye for design > fast back-and-forth, it runs 30%+ faster • keep it for Opus 5.5 > long, open-ended work with judgment calls for hours > anything where one wrong assumption costs you a day • the setting that saves the most > Medium is the default in the Claude apps and Claude Code > on Low or Medium it beats Sonnet 5's best score for a tenth of the cost > raise effort only when the first answer falls short • the price > $2 in and $10 out, against $4 and $20 for Opus > cache reads $0.20 on both • check before you switch > if you ran Sonnet with thinking off, move to the new between_tools setting first whoever routes the weekly grunt work to Sonnet 5.5 and saves Opus for the hard calls ships more on half the budget, and that's the person who gets noticed
Annatar.md19,642 görüntüleme • 28 gün önce
CLAUDE + OBSIDIAN + KARPATHY'S LLM WIKI = A WIKI THAT MAINTAINS ITSELF run this once and your research compounds instead of evaporating sources go in, a wiki comes out, and a loop keeps it honest • three layers, one owner each > the schema - CLAUDE.md, you and the agent co-evolve it > the wiki - markdown the agent owns. it writes, you read > raw sources - immutable, read only, never edited • three operations on repeat > ingest - read the source, write the summary, log one line > query - answer with citations, file it back as a new page > lint - contradictions, orphans and dead links caught early RAG re-derives the same answer from raw chunks on every query. you pay for that thinking again, every time you ask this wiki gets written once and stays current. 10-15 pages per source, nearly 100 sources on the index alone month one it saves you an hour a day. month six it knows your field better than the person you're up against for the senior seat "Obsidian is the IDE, the LLM is the programmer, the wiki is the codebase" - Andrej Karpathy the stack: 📁 Obsidian ↳ 📁 Claude ↳ 📁 Karpathy LLM-Wiki ↳
Annatar.md16,939 görüntüleme • 1 ay önce