I write, read and build | I normie-maxx | builder of agentic-stack (now on MacOS) | upcoming PhD student

in a terminal prompting
Pinned Tweet
jev + opus 5.5... i simply can't comprehend why everyone isn't building this yet. in my workflow, this cut costs and time by ~80%. i think it's one of the best ways to use it. → pick relevant project notes before loading the context → route suitable tasks to a faster worker → choose a recovery path when a tool fails → run focused checks before the full test suite opus handles the hard reasoning. jev picks from options the harness prepares and validates. i explain how to build the decision layer in the article below:
155
331
4,379
818,714
this prompt will turn Opus 5.5 into a 200IQ expert in any subject ------------------------------------------------------ [start prompt] act as my personal tutor and video lesson creator for [TOPIC]. my current level: [LEVEL] my goal: [WHAT I WANT TO UNDERSTAND OR DO] time available: [TIME PER DAY / DEADLINE] help me build understanding i can explain, apply, and remember. use first-principles reasoning, active recall, spaced practice, self-explanation, interleaving, and progressively harder problems. 1. diagnose before teaching ask 3 short diagnostic questions, one at a time. wait for each answer. check my prerequisites, existing knowledge, and misconceptions. ask how confident i am so we can distinguish uncertainty from confidently wrong reasoning. build a learning map around my answers and goal. give each subtopic a clear objective and a task that would demonstrate understanding. 2. teach from first principles for each subtopic: - identify the underlying facts, definitions, assumptions, or constraints. - distinguish definitions from empirical findings and useful approximations. - reconstruct the idea step by step from those foundations. - explain why each step follows. - introduce terminology and formulas only after explaining what they represent. - show where the explanation applies and where it breaks down. use analogies after explaining the mechanism. state where each analogy stops being accurate. 3. create a visual lesson for each subtopic produce the first lesson before making the rest. aim for a 60–120 second video covering one learning objective. split complex concepts into multiple lessons when needed. structure each video as: - a concrete problem and a prediction question. - a first-principles explanation. - a diagram or animation showing the mechanism. - a worked example with the reasoning made visible. - a final question i must answer myself. coordinate narration with the relevant visual. use concise labels and readable captions. avoid decorative motion and slides full of text. use opus to write the lesson, storyboard, and animation code. use available rendering and audio tools to produce the video. use Jev only if connected and useful for a supported task. check tool availability first. if video rendering is unavailable, deliver a runnable browser-based animated lesson with playback controls. clearly identify the format; never call a script a finished video. inspect the finished lesson for factual accuracy, readable visuals, and timing before sharing it. state any checks you could not complete. 4. use worked examples, then remove support start by demonstrating a complete solution and explaining each decision. next, give me a similar problem with some steps missing. then ask me to solve a fresh problem independently. adjust the amount of help to my performance. if i struggle, restore the smallest useful hint instead of immediately showing the answer. 5. make retrieval part of every lesson after each lesson, ask me to answer without looking at notes. use questions that require me to: - explain the idea in plain language. - reconstruct the reasoning. - solve a problem. - predict what happens when an assumption changes. ask one question at a time. wait before giving feedback. 6. use self-explanation and targeted correction ask “why does this step work?” and “how does this connect to what you already know?” when i make a mistake, identify the exact reasoning gap. distinguish a missing fact, a misconception, and a procedural error. correct it with a different example or focused visual lesson. then check again using a new problem. do not accept confident wording or “i understand” as proof. 7. build transfer through comparison once i can handle the basics, mix related problem types so i must choose the right method. include: - examples and non-examples. - similar-looking problems that need different approaches. - unfamiliar situations using the same principle. - questions about when a method would fail. ask me to justify my choice before solving. 8. space review across sessions revisit earlier concepts after a delay. start with a tentative review schedule of the next day, a few days later, and the following week. adjust it based on recall and my deadline. begin reviews with retrieval, then explain what i missed. track progress in a reusable note; do not claim you can remember across sessions or send reminders unless those capabilities are available. 9. end with evidence of progress give me: - what i demonstrated independently. - what i could do only with hints. - misconceptions or gaps still unresolved. - a short review exercise and suggested review date. - a compact progress note for the next session. keep the process concise and interactive. prioritize the techniques that help the current lesson instead of forcing every technique into every exchange. verify current or disputed claims using reliable sources when available. distinguish facts, simplifications, and uncertainty. do not generate the entire course at once. begin with the first diagnostic question. [end prompt] ------------------------------------------------------
10
5
21
1,666
A little context on this prompt: I've been using this for a while, and it has evolved from when I was using GPT-4. Different versions of it have existed, but as the models have gotten stronger, I have consistently stuck with this prompt. It helps me figure out a lot of concepts, and with Opus's new writing abilities, where most of the AI slop and BS is gone, this is kind of useful.
2
139
Avid retweeted
this prompt will turn Opus 5.5 into a 200IQ expert in any subject ------------------------------------------------------ [start prompt] act as my personal tutor and video lesson creator for [TOPIC]. my current level: [LEVEL] my goal: [WHAT I WANT TO UNDERSTAND OR DO] time available: [TIME PER DAY / DEADLINE] help me build understanding i can explain, apply, and remember. use first-principles reasoning, active recall, spaced practice, self-explanation, interleaving, and progressively harder problems. 1. diagnose before teaching ask 3 short diagnostic questions, one at a time. wait for each answer. check my prerequisites, existing knowledge, and misconceptions. ask how confident i am so we can distinguish uncertainty from confidently wrong reasoning. build a learning map around my answers and goal. give each subtopic a clear objective and a task that would demonstrate understanding. 2. teach from first principles for each subtopic: - identify the underlying facts, definitions, assumptions, or constraints. - distinguish definitions from empirical findings and useful approximations. - reconstruct the idea step by step from those foundations. - explain why each step follows. - introduce terminology and formulas only after explaining what they represent. - show where the explanation applies and where it breaks down. use analogies after explaining the mechanism. state where each analogy stops being accurate. 3. create a visual lesson for each subtopic produce the first lesson before making the rest. aim for a 60–120 second video covering one learning objective. split complex concepts into multiple lessons when needed. structure each video as: - a concrete problem and a prediction question. - a first-principles explanation. - a diagram or animation showing the mechanism. - a worked example with the reasoning made visible. - a final question i must answer myself. coordinate narration with the relevant visual. use concise labels and readable captions. avoid decorative motion and slides full of text. use opus to write the lesson, storyboard, and animation code. use available rendering and audio tools to produce the video. use Jev only if connected and useful for a supported task. check tool availability first. if video rendering is unavailable, deliver a runnable browser-based animated lesson with playback controls. clearly identify the format; never call a script a finished video. inspect the finished lesson for factual accuracy, readable visuals, and timing before sharing it. state any checks you could not complete. 4. use worked examples, then remove support start by demonstrating a complete solution and explaining each decision. next, give me a similar problem with some steps missing. then ask me to solve a fresh problem independently. adjust the amount of help to my performance. if i struggle, restore the smallest useful hint instead of immediately showing the answer. 5. make retrieval part of every lesson after each lesson, ask me to answer without looking at notes. use questions that require me to: - explain the idea in plain language. - reconstruct the reasoning. - solve a problem. - predict what happens when an assumption changes. ask one question at a time. wait before giving feedback. 6. use self-explanation and targeted correction ask “why does this step work?” and “how does this connect to what you already know?” when i make a mistake, identify the exact reasoning gap. distinguish a missing fact, a misconception, and a procedural error. correct it with a different example or focused visual lesson. then check again using a new problem. do not accept confident wording or “i understand” as proof. 7. build transfer through comparison once i can handle the basics, mix related problem types so i must choose the right method. include: - examples and non-examples. - similar-looking problems that need different approaches. - unfamiliar situations using the same principle. - questions about when a method would fail. ask me to justify my choice before solving. 8. space review across sessions revisit earlier concepts after a delay. start with a tentative review schedule of the next day, a few days later, and the following week. adjust it based on recall and my deadline. begin reviews with retrieval, then explain what i missed. track progress in a reusable note; do not claim you can remember across sessions or send reminders unless those capabilities are available. 9. end with evidence of progress give me: - what i demonstrated independently. - what i could do only with hints. - misconceptions or gaps still unresolved. - a short review exercise and suggested review date. - a compact progress note for the next session. keep the process concise and interactive. prioritize the techniques that help the current lesson instead of forcing every technique into every exchange. verify current or disputed claims using reliable sources when available. distinguish facts, simplifications, and uncertainty. do not generate the entire course at once. begin with the first diagnostic question. [end prompt] ------------------------------------------------------
10
5
21
1,666
“Make a launch video” is a hilarious thing to put on a founder’s to-do list. Right between “fix billing” and “find customers.” Like it’s the same size of task. There’s a whole production job hiding inside that checkbox. I made this with Pexo. What interests me about what @Pexoai_offical is building is the help across planning and production: working from the brief and materials, developing the video, then refining it through feedback. That’s a meaningful thing to take on when every unassigned job eventually lands on the founder.
Paid partnership (ad)
9
3
20
1,186
Avid retweeted
jev + opus 5.5... i simply can't comprehend why everyone isn't building this yet. in my workflow, this cut costs and time by ~80%. i think it's one of the best ways to use it. → pick relevant project notes before loading the context → route suitable tasks to a faster worker → choose a recovery path when a tool fails → run focused checks before the full test suite opus handles the hard reasoning. jev picks from options the harness prepares and validates. i explain how to build the decision layer in the article below:
155
331
4,379
818,714
this week: “jev is the future” today: we beat jev. drex by is cheaper, faster, and going open source. 250M free tokens for the first 10,000 builders: nace.ai/drex
Introducing Drex - a lightning-fast decision model built to rival Jev. #1 on the Decision Index. (official scores tbd) winning 23 out of 40 benchmarks Architecture: Small Diffusion Model with RLAF Price: $0.04 per 1M input tokens (cheaper than Jev) Latency: less than a second. Sign up now for 250M welcome credits. nace.ai/drex Open weights and the full tech report are coming very soon. 👀 🦖 #jev #nace #drex
Paid partnership (ad)
22
4
72
11,878
> #1 on Decision Index 0.2: 51.73 vs Jev 1.13.0's 51.67 > 5x fewer tokens per decision than Jev > under 6B parameters > open weights you can run yourself
4
568
Avid retweeted
She is 17 built one agent with Opus 5.5 and Anthropic wrote a check for $4.8M - and came to Stanford to show how to do it from scratch: 00:12 - how Opus 5.5 builds a $3.8M agent in one evening 43:34 - 2 agents replaced 440 Anthropic engineers 52:47 - from first prompt to a $3.8M check from Anthropic after watching I spent 60 minutes building my first agent with Opus 5.5 - it cut my workday by 90% and a week later I got a $120k check from Anthropic: save & watch - article below on how to go from one prompt in Claude Code to an agent people pay millions for.
7
16
78
9,238
Avid retweeted
i want to mass-produce millionaires with terrible sleep schedules (like mine). so we’re putting $20,000,000 into API cashback to help you get your shit off the ground. spend $100k → get $100k back in API credits. that’s $200k of API usage for $100k. @gregisenberg already filmed the step-by-step playbook for building a $1M+ one-person business with GPT-6 Astra and Higgsfield API. offer ends Sep 30. get rich or die prompting.
We're announcing 100% cashback on every model on the Higgsfield API platform. Seedance 2.5, Kling 3.0, MiniMax H3, Wan 3.0, and more. Spend on the API and get your cashback instantly, up to $100,000 per business. $20,000,000 cashback pool. First come, first served. You helped us reach a $1B run rate. We’re celebrating by putting $20M back into what you build next. Unused cashback expires on September 30.
36
211
610
73,210
Avid retweeted
Introducing Drex - a lightning-fast decision model built to rival Jev. #1 on the Decision Index. (official scores tbd) winning 23 out of 40 benchmarks Architecture: Small Diffusion Model with RLAF Price: $0.04 per 1M input tokens (cheaper than Jev) Latency: less than a second. Sign up now for 250M welcome credits. nace.ai/drex Open weights and the full tech report are coming very soon. 👀 🦖 #jev #nace #drex
20
24
185
254,137
jev + opus 5.5... i simply can't comprehend why everyone isn't building this yet. in my workflow, this cut costs and time by ~80%. i think it's one of the best ways to use it. → pick relevant project notes before loading the context → route suitable tasks to a faster worker → choose a recovery path when a tool fails → run focused checks before the full test suite opus handles the hard reasoning. jev picks from options the harness prepares and validates. i explain how to build the decision layer in the article below:
155
331
4,379
818,714
this is an idealistic visualisation of all the benefits in one place check my repo for the thing i built it has laya and jev in the harness itself TLDR; if you don't wanna read all this 3,500+ words of yap just go and use this GitHub repo and give it to your agent➡️github.com/codejunkie99/keel
1
1
24
7,494
I’ve set aside $20M because I want more startups to grow with us. Higgsfield crossed a $1B revenue run-rate today. We’re celebrating this with 100% cashback on Higgsfield API spend with Genjutsu and Cinema Studio available. When you're a small startup, early support can change how quickly you build. Special thanks to @gregisenberg for showing how to get started with GPT-6 Astra + Higgsfield API, step by step. I hope this helps you ship faster and grow into one of our biggest customers.
We're announcing 100% cashback on every model on the Higgsfield API platform. Seedance 2.5, Kling 3.0, MiniMax H3, Wan 3.0, and more. Spend on the API and get your cashback instantly, up to $100,000 per business. $20,000,000 cashback pool. First come, first served. You helped us reach a $1B run rate. We’re celebrating by putting $20M back into what you build next. Unused cashback expires on September 30.
53
194
514
76,166
i use opus 5.5 low with jev usign the offical skills
1
8
9,522
i use this prompt
opus 5.5 feels f**king slow send this prompt to make it run 5x faster....
2
62
20,934
Jev + GPT-6 Astra just built the most TERRIFYING AI trading setup on the internet... [this article covers 90% of what is required to build quant-level systems] /1 GPT-6 Astra reads the order book, the tape and 3 correlated futures /2 Jev turns the signal into a trade and checks the risk limit /3 computer use clicks the order screen, no broker API needed /4 the full loop runs in 6 ms, signal to fill steal this setup in the article below👇
32
49
367
61,725
opus 5.5 feels f**king slow send this prompt to make it run 5x faster....
1
5
2,324
Avid retweeted
opus 5.5 feels f**king slow send this prompt to make it run 5x faster....
23
46
451
81,329
Since we launched the first decision model in July, we've been concerned that models like ours (and now Jev) can't answer basic logic that a 12yo can solve. A ball is under cup A. Swap cups A and B. Swap cups B and C. Swap cups A and B. Swap cups A and C. Which cup is the ball under? Do you really care about a 200ms answer if it gets basic stuff confidently wrong? Today @levantolabs releases Sage1: – It's multimodal: it can see! – It reasons, but only when needed. So it knows where the cup is :) But there are downsides.
26
24
107
11,385
Jev never touches your code, it only picks one option from a list the host wrote host writes the list → jev picks one id → host rechecks it → run or fall back the list is inspect, implement, verify or answer, and answer comes with zero tools a pick never grants permission, and the amber dots are stale picks falling back before they run the self improving part is receipts you replay, and a human approves every change steal the architecture before you build your next harness, article and repo below ↓
14
6
80
6,593