Annatar.md

I have seen many fail with crude prompts
Let me teach you the hidden ways of Claude

Skills, agents, and lectures | @aiscwork partner

[Annatar.md @AnnatarXBT]

ANDREJ KARPATHY WAS RIGHT. THIS 40-MINUTE AI LECTURE PROVES IT

we're in the 1960s of LLMs
most people using Claude have the wrong mental model

Software 3.0. LLMs as operating systems
how to actually think about and work with them

save this 👇

[Mr. Buzzoni @polydao]

https://t.co/rSl7yKpZy7

[Annatar.md @AnnatarXBT]

THE CLAUDE CODE TEAM WROTE THE FIELD GUIDE ON HANDING YOUR OWN WORK OFF TO LOOPS

a loop is an agent repeating cycles until a stop condition is met

there are 4. each one hands off one more piece of the job than the last:

> turn-based - you hand off the check
Claude verifies its own work through a SKILL.md

> goal-based - you hand off the stop condition
/goal sets the target, an evaluator sends Claude back until it clears "Lighthouse 90, stop after 5 tries"

> time-based - you hand off the trigger
/loop runs every 5 min to check the PR and fix CI. /schedule moves it to the cloud

> proactive - you hand off the prompt itself
a routine watches a channel, triages, fixes, and a second agent reviews the work with no human there

the reframe: a loop is only as good as the system you build around it

pick one task where you are the bottleneck
hand off the first piece tonight

save this 👇

[CyrilXBT @cyrilXBT]

https://t.co/lkgDX3DkdY

[Annatar.md @AnnatarXBT]

@polydao this sounds way more useful than my usual note setup tbh

[Annatar.md @AnnatarXBT]

@rimtoln sell the result not the roadmap

most people are stuck selling vision when they should be shipping

[Annatar.md @AnnatarXBT]

> do you understand what just happened to the job market
> one person + Claude Skills
> does the work of an entire dev team
> $20/month vs $200K/year in salaries
> the Google engineer who automated 80% of his job?
> he built it straight from Anthropic's own skills guide
> i read the whole thing so you don't have to
> bookmark this before you waste another token 👇

[Mr. Buzzoni @polydao]

https://t.co/rSl7yKpZy7

[Annatar.md @AnnatarXBT]

@irongiantXBT karpathy is a legend

[Annatar.md @AnnatarXBT]

@polydao bookmarked

[Annatar.md @AnnatarXBT]

@polydao been manually watching congress trades for months, didn't know i could just automate this whole thing

[Annatar.md @AnnatarXBT]

TYPES OF LOOPS - FOUR WAYS TO STRUCTURE AGENTIC WORK WITH CLAUDE

loop engineering gets talked about like one thing -> it's really four structures, and picking the right one is the whole skill

every loop answers two questions: what starts a run, and what decides it's done. by hand you answer both yourself

each type moves more of that into the system

1) turn-based
> starts on your prompt
> Claude gathers context, acts, checks itself, replies - then you write the next one
best for: requirements still forming

2) goal-based
> starts with /goal + a limit ("Lighthouse 90, stop after 5 tries")
> when Claude tries to stop, an evaluator checks the goal and sends it back on a no
best for: measurable outcomes where the path isn't worth watching

3) time-based
> starts on a clock ("/loop every 5 min: check the PR, fix CI")
> /loop runs local, /schedule moves it to the cloud
best for: recurring work you can define ahead of time

4) proactive
> starts on an event, no human present
> /schedule watches a channel, spawns triage, fix, and a reviewer that judges the work before it closes
best for: standing duties where you can't predict what comes in

each type hands off one more job than the last. the more you hand off, the less you babysit

full breakdown on loop engineering in the article below 👇

[Morty @0xMortyx]

https://t.co/pRcdgBOukk

[Annatar.md @AnnatarXBT]

@polydao chat with the content after transcription is the feature i didn't know i was missing

[Annatar.md @AnnatarXBT]

SKIP NETFLIX TONIGHT. GIVE THIS ONE HOUR INSTEAD

> it's a full course on Claude AI - how to build with it, how to automate your actual work with it, start to finish

people who sit through this tonight wake up with a skill most people still won't have two years from now

> people who scroll past it will be back on Netflix next year asking why nothing changed

and with Fable 5 access extended through July 19, this is also your window to test the heaviest parts of the course on the strongest model Anthropic has

your call

[Codez @0xCodez]

https://t.co/eJB3TgyJwV

[Annatar.md @AnnatarXBT]

ANTHROPIC ITSELF JUST PROVED HOW BADLY PEOPLE PROMPT CLAUDE

> their applied AI team dropped a free 22-minute workshop

not a creator guessing from the outside - the actual team that wrote the weights

> here's the part that stings: a properly structured Claude prompt runs on 6 elements, and most of us are stuck using 1, maybe 2 if we're trying

that single gap explains the answer that felt slightly off, the four rewrites, the prompt that worked once and never again

> watch it tonight and you know something most daily users never will -> skip it and you keep paying full price for a fraction of the tool

i watched it twice. then built a Claude Skill that runs all 6 elements on every prompt by itself, so i stopped thinking about structure completely

workshop first, skill setup below. this one compounds 👇

[Annatar.md @AnnatarXBT]

THIS "GETTING STARTED WITH LOOPS" DOC TURNS CLAUDE INTO AN EMPLOYEE THAT SHIPS WHILE YOU SLEEP

spoiler: prompt engineering didn't survive

the 4 loop types, from you-drive to fully hands-off:

> turn-based - 25% autonomy. one task per round, control returns to you. best for exploration when requirements are unclear

> goal-based - 70%. you define "done" with /goal and a separate evaluator keeps sending Claude back until it clears the bar. tests passing, target scores, bugs cleared

> time-based - 60%. /loop fires on a schedule: check the PR every 5 minutes, address review comments, fix failing CI

> proactive - 95%. an event fires, a router dispatches child agents in parallel, a review model oversees the output. no human online at all

the advice in it that matters most: loop quality depends on system quality, and tokens are the real cost

knowing the 4 types is level one

this article is the full 14-step roadmap from typing prompts by hand to designing loops - /goal, SKILL.md, worktrees, MCP, all of it

start climbing 👇

[Annatar.md @AnnatarXBT]

@zefirium thanks for sharing - more content like this coming

[Annatar.md @AnnatarXBT]

THIS "GETTING STARTED WITH LOOPS" DOC TURNS CLAUDE INTO AN EMPLOYEE THAT SHIPS WHILE YOU SLEEP

spoiler: prompt engineering didn't survive

the 4 loop types, from you-drive to fully hands-off:

> turn-based - 25% autonomy. one task per round, control returns to you. best for exploration when requirements are unclear

> goal-based - 70%. you define "done" with /goal and a separate evaluator keeps sending Claude back until it clears the bar. tests passing, target scores, bugs cleared

> time-based - 60%. /loop fires on a schedule: check the PR every 5 minutes, address review comments, fix failing CI

> proactive - 95%. an event fires, a router dispatches child agents in parallel, a review model oversees the output. no human online at all

the advice in it that matters most: loop quality depends on system quality, and tokens are the real cost

knowing the 4 types is level one

this article is the full 14-step roadmap from typing prompts by hand to designing loops - /goal, SKILL.md, worktrees, MCP, all of it

start climbing 👇

[Codez @0xCodez]

https://t.co/eJB3TgyJwV

[Annatar.md @AnnatarXBT]

@irongiantXBT bookmarked

[Annatar.md @AnnatarXBT]

Claude Code will grill you with 40+ questions before writing a single line of code

> it's called /grill-me - 3 sentences long. most impactful skill I use

instead of jumping straight to code - it walks every branch of your design tree until there's zero ambiguity

> every question reveals something you hadn't thought of

other skills I use daily:

> /write-a-prd - idea -> proper product doc

> /prd-to-issues - doc -> GitHub issues automatically

> /tdd - tests first, forces edge case thinking before code

> /improve-codebase-architecture -> full structural review of your codebase

full video from Matt Pocock 👇

[Khairallah AL-Awady @eng_khairallah1]

https://t.co/mtiWDoKy6m

[Annatar.md @AnnatarXBT]

@polydao bookmarked

[Annatar.md @AnnatarXBT]

I ABANDONED LOVABLE AFTER SEEING WHAT CLAUDE CODE CAN ACTUALLY DO

basic generators are great for quick ideas
but trying to build a premium site with Lovable feels like playing in a sandbox

i switched to Claude Code running on the Fable 5 reasoning engine:

> local execution - the agent reads and edits files directly inside the workspace
> advanced reasoning - Fable 5 writes complex styling instead of generic templates
> high speed - it built a premium interface worth $8,260 in less than 7 minutes

you get raw, uninhibited codebase access instead of clicking through a limited web editor

grab the full guide below 👇

[Asteri @Asteri_eth]

https://t.co/V3U5aEPJ66

[Annatar.md @AnnatarXBT]

@polydao minimax ceo is simply the best

[Annatar.md @AnnatarXBT]

do you understand what Claude just did?

anthropic is preventing knowledge workers from losing original thinking skills

the dashboard acts as a personal clinical auditor for your productivity:

> task delegation - isolates writing and coding tasks you should keep human
> prompt tracking - flags vague requests to fix your instruction habits
> mental limits - schedules automatic reminders to stop chat sessions
> data protection - runs audits without index-syncing sensitive inbox files

you protect your original thinking while letting the machine run loops

what is the one task you will never delegate to Claude, even if it runs 10x faster?

[Claude @claudeai]

Introducing a new way to reflect on how you use Claude.

Your monthly recap shows when you use Claude most and what you spent that time working on, with options to set quiet hours and nudges to take breaks. Find your dashboard in Settings under Reflect: https://t.co/8QAn47W5rI