Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

I reverse-engineered Claude Code's binary to add a feature I always wanted: When context fills up - instead of nuking everything with /compact - I can now surgically strip tool calls/results and thinking blocks while keeping all actual messages intact.

Convergence Boy

2,819 subscribers

197,974 views • 4 months ago •via X (Twitter)

Science & Technology

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

OMG I can't believe I reverse engineered Apple Notes with Claude Code. I can import everything from Apple Notes like folders, notes, images and everything. Also this is realtime, not sped up. Now everyone can migrate from Apple Notes to Grape.

OMG I can't believe I reverse engineered Apple Notes with Claude Code. I can import everything from Apple Notes like folders, notes, images and everything. Also this is realtime, not sped up. Now everyone can migrate from Apple Notes to Grape.

Ozgur Ozer

173,126 views • 4 months ago

I reverse-engineered Claude Code's binary. Found a flag they hid from --help. --sdk-url Enable it and the terminal disappears. The CLI becomes a WebSocket client. We built a server to catch the connection. Added a React UI on top. Now I run Claude Code from my browser. From my phone. From anywhere. Same $200/month subscription. Zero extra API costs. bunx the-vibe-companion

I reverse-engineered Claude Code's binary. Found a flag they hid from --help. --sdk-url Enable it and the terminal disappears. The CLI becomes a WebSocket client. We built a server to catch the connection. Added a React UI on top. Now I run Claude Code from my browser. From my phone. From anywhere. Same $200/month subscription. Zero extra API costs. bunx the-vibe-companion

Stan Girard

693,421 views • 5 months ago

While working on my new Goku SSJ Mid I thought that maybe I can do what I've always wanted to with a SSJ hair and turn it gold when charging KI, well, since we can actually add code to the BPs now, I managed to do just that and change materials when Charging or in Sparking mode

While working on my new Goku SSJ Mid I thought that maybe I can do what I've always wanted to with a SSJ hair and turn it gold when charging KI, well, since we can actually add code to the BPs now, I managed to do just that and change materials when Charging or in Sparking mode

Millennium

259,530 views • 1 year ago

I've been writing #rstats packages and coding web maps for a decade now. Despite this, while I've always wanted a story map feature in R, I never really knew how. With Claude Sonnet and Cursor, I got a working prototype in a morning, and now it's on CRAN for you to use.

I've been writing #rstats packages and coding web maps for a decade now. Despite this, while I've always wanted a story map feature in R, I never really knew how. With Claude Sonnet and Cursor, I got a working prototype in a morning, and now it's on CRAN for you to use.

Kyle Walker

21,259 views • 1 year ago

Ranamon is one of those characters I wanted to dedicate an entire animation to. Mostly because of those amazing feet. It has been on my plate for quite a while now and I can proudly say that I'm satisfied with the results.

Ranamon is one of those characters I wanted to dedicate an entire animation to. Mostly because of those amazing feet. It has been on my plate for quite a while now and I can proudly say that I'm satisfied with the results.

BlueHulkhog

17,818 views • 2 years ago

Vibe coded this Threejs avatar gallery with over 300 3D open source avatars in my spare time. Cursor + Claude doing all of the heavy lifting while I watched Features a complete VRM inspector tool and Japanese support! Probably over engineered but works

ToxSam

22,174 views • 1 year ago

I'm SO impressed by Claude Skills. They are the best way to package custom instructions (and code) for the agent to access. (I prefer this to Claude Code's subagents because of the way it handles context) I used the skill-creator skill (comes default inside Claude) to create a skill for designing a presentation like Steve Jobs or Seth Godin, then used that skill to create a PPT about Claude Code. Below is a walkthrough of the results. The output when using the above mentioned skill vs when i don't is so clear. I'm bullish on what this implies for both Claude web and Claude Code. Extremely powerful. Anthropic cooked! 🔥

I'm SO impressed by Claude Skills. They are the best way to package custom instructions (and code) for the agent to access. (I prefer this to Claude Code's subagents because of the way it handles context) I used the skill-creator skill (comes default inside Claude) to create a skill for designing a presentation like Steve Jobs or Seth Godin, then used that skill to create a PPT about Claude Code. Below is a walkthrough of the results. The output when using the above mentioned skill vs when i don't is so clear. I'm bullish on what this implies for both Claude web and Claude Code. Extremely powerful. Anthropic cooked! 🔥

Nityesh

62,918 views • 9 months ago

you're running /compact every 10 minutes to fix a problem that disappears the second you give Claude one job instead of five. Anthropic shipped a clean 4-minute video on context management. /compact when the window fills up. /clear to reset. small files over big files. the practical tips work. the video covers the cleanup. it doesn't cover why your window keeps filling up in the first place, and the answer is usually that you're holding too many jobs in one prompt. look at what most people put in a single Claude call: > guess the right answer > choose between options > set the position size > watch the exit window > read the news for sentiment five tools' worth of state in one conversation. of course the window fills. i ran into this on a Polymarket bot: the prediction-layer version ate context constantly. /compact every 8-10 turns. i thought i needed better memory hygiene. i needed fewer jobs per call. i stripped the prompt down to one job: veto an entry i had already decided to take. the context window stopped filling. /compact went from every 8 turns to once a session, if ever. over 90 days that prompt blocked 128 trades. watch the 4 min, then ask why your window keeps needing it. clean context. lean prompt. nothing to compact.

you're running /compact every 10 minutes to fix a problem that disappears the second you give Claude one job instead of five. Anthropic shipped a clean 4-minute video on context management. /compact when the window fills up. /clear to reset. small files over big files. the practical tips work. the video covers the cleanup. it doesn't cover why your window keeps filling up in the first place, and the answer is usually that you're holding too many jobs in one prompt. look at what most people put in a single Claude call: > guess the right answer > choose between options > set the position size > watch the exit window > read the news for sentiment five tools' worth of state in one conversation. of course the window fills. i ran into this on a Polymarket bot: the prediction-layer version ate context constantly. /compact every 8-10 turns. i thought i needed better memory hygiene. i needed fewer jobs per call. i stripped the prompt down to one job: veto an entry i had already decided to take. the context window stopped filling. /compact went from every 8 turns to once a session, if ever. over 90 days that prompt blocked 128 trades. watch the 4 min, then ask why your window keeps needing it. clean context. lean prompt. nothing to compact.

Mnimiy

20,635 views • 2 months ago

Winter’s ment about ‘Blue’ “This is a song I wrote bc I wanted to bring comfort to MYs. So when you feel like you’ve lost a bit of confidence or when you need comfort, I hope you can listen to this song while thinking that I’m the one cheering u up and comforting you” 🥹

Winter’s ment about ‘Blue’ “This is a song I wrote bc I wanted to bring comfort to MYs. So when you feel like you’ve lost a bit of confidence or when you need comfort, I hope you can listen to this song while thinking that I’m the one cheering u up and comforting you” 🥹

Jazz

39,632 views • 4 months ago

I FINALLY FIGURED OUT WHY CLAUDE KEEPS FORGETTING THINGS. For months, I thought Claude’s memory was broken. I’d spend 20 minutes explaining my project few chats later, it would forget important stuff and remember something completely random. Turns out the problem wasn’t Claude. It was how I was using it. Here’s the system that fixed it LEVEL 1: Clean up the basics (5 mins) - Remove old memories you don’t need - Add proper project instructions - Tell Claude exactly what should and shouldn’t be remembered Most people never do this and then wonder why memory feels inconsistent. LEVEL 2: Build a context system (30-60 mins) Create a folder with: → Instructions.md → Memory.md → Context.md → Archive.md Now Claude has a dedicated place to store information instead of trying to remember everything from chat history. Massive difference. LEVEL 3: Build a real AI second brain → Connect Claude to your notes. → Research. → Documents. → Knowledge base. Instead of starting from zero every session, it builds context over time. This is when Claude stops feeling like a chatbot and starts feeling like an actual assistant. Most people are obsessed with prompts. The real power comes from building better context.

I FINALLY FIGURED OUT WHY CLAUDE KEEPS FORGETTING THINGS. For months, I thought Claude’s memory was broken. I’d spend 20 minutes explaining my project few chats later, it would forget important stuff and remember something completely random. Turns out the problem wasn’t Claude. It was how I was using it. Here’s the system that fixed it LEVEL 1: Clean up the basics (5 mins) - Remove old memories you don’t need - Add proper project instructions - Tell Claude exactly what should and shouldn’t be remembered Most people never do this and then wonder why memory feels inconsistent. LEVEL 2: Build a context system (30-60 mins) Create a folder with: → Instructions.md → Memory.md → Context.md → Archive.md Now Claude has a dedicated place to store information instead of trying to remember everything from chat history. Massive difference. LEVEL 3: Build a real AI second brain → Connect Claude to your notes. → Research. → Documents. → Knowledge base. Instead of starting from zero every session, it builds context over time. This is when Claude stops feeling like a chatbot and starts feeling like an actual assistant. Most people are obsessed with prompts. The real power comes from building better context.

nordin.eth

50,500 views • 1 month ago

Pi was built when there were already agent harnesses around. Here’s why Mario Zechner(Mario Zechner), found them suboptimal and built Pi, a minimalist self-modifying agent: #1 - Mario initially was a believer in Claude Code: "I was a believer in Claude code because they were the first that packaged agentic search up in a really compelling package. And at the time that fit my workflow really well. Everything around the LLM was kind of nice and tidy and easy to understand. I was super happy. I was proselytising Claude code." #2 - Reverse engineering Claude Code highlighted the degradation that Mario felt as a user: "I personally like simple tools that are stable and that I can rely on. Even if they have non-deterministic parts, all the deterministic parts should be as stable as possible. That was just not the experience with Claude Code around summer 2025. They would take away your control of the context. They would inject stuff behind your back, which is bad. Then, your workflows stopped working because there's now a system reminder that you don't even see in the UI that would modify the behaviour of the model. They would also do this to the system prompt. I built a little service where I can track the progression or evolution of the system, prompt and tool definitions and, with every release, it was messing with stuff. That just messed with my workflows and I don't appreciate that." #3 - PI was built with an appreciation for simple and reliable tools: "If I commit to a development tool, I want it to be a stable, reliable thing like a hammer. I don't want my hammer to break a different spot every day. That's terrible. We need somebody who goes the full velocity kind of way. But I don't want to work with a tool like that."

Pi was built when there were already agent harnesses around. Here’s why Mario Zechner(Mario Zechner), found them suboptimal and built Pi, a minimalist self-modifying agent: #1 - Mario initially was a believer in Claude Code: "I was a believer in Claude code because they were the first that packaged agentic search up in a really compelling package. And at the time that fit my workflow really well. Everything around the LLM was kind of nice and tidy and easy to understand. I was super happy. I was proselytising Claude code." #2 - Reverse engineering Claude Code highlighted the degradation that Mario felt as a user: "I personally like simple tools that are stable and that I can rely on. Even if they have non-deterministic parts, all the deterministic parts should be as stable as possible. That was just not the experience with Claude Code around summer 2025. They would take away your control of the context. They would inject stuff behind your back, which is bad. Then, your workflows stopped working because there's now a system reminder that you don't even see in the UI that would modify the behaviour of the model. They would also do this to the system prompt. I built a little service where I can track the progression or evolution of the system, prompt and tool definitions and, with every release, it was messing with stuff. That just messed with my workflows and I don't appreciate that." #3 - PI was built with an appreciation for simple and reliable tools: "If I commit to a development tool, I want it to be a stable, reliable thing like a hammer. I don't want my hammer to break a different spot every day. That's terrible. We need somebody who goes the full velocity kind of way. But I don't want to work with a tool like that."

The Pragmatic Engineer

62,825 views • 2 months ago

I got curious how compaction works as a PM, so I did some brain surgery on Claude Code: (Anthropic's been doing really interesting work on context editing - they showed Claude Opus playing Settlers of Catan for 75+ minutes in a single thread by constantly editing the context instead of starting fresh. When I saw that Claude Code has a compaction command with optional custom instructions, I wanted to understand what's actually happening.) Abhishek Katiyar and Aman Khan gave me the key tip: Claude Code stores all your conversation history as text files on your computer. Open a new directory and give Claude Code a task. Here's how to watch compaction happening: 1. Go to your user's root directory 2. Press Command+Shift+Period (Mac) to show hidden folders 3. Navigate to ~/.claude/projects/ 4. Find your project folder and use Cursor/VSCode to open it (there's a reason) 5. Install the JSONL Gazelle plugin (open source, thank you Gabor Cselle!) 6. Open the most recent JSONL file - each row is a message in your conversation 7. Run the compact command in Claude Code with custom instructions 8. Watch what happens in the file What I learned: When you compact, Claude Code doesn't just summarize and delete everything. It creates a "compact boundary" in the conversation file, writes a summary of what happened before, but keeps the full original conversation (!!!!) The new thread can still retrieve any details from before compaction if needed. That is so damn cool. Why this matters: What you're getting in Claude Code is similar to what Anthropic ships in their developer SDK - so inspecting your daily tools is how you build real product intuition. The best way to understand AI systems is to open them up and look inside. Everything is text files.

I got curious how compaction works as a PM, so I did some brain surgery on Claude Code: (Anthropic's been doing really interesting work on context editing - they showed Claude Opus playing Settlers of Catan for 75+ minutes in a single thread by constantly editing the context instead of starting fresh. When I saw that Claude Code has a compaction command with optional custom instructions, I wanted to understand what's actually happening.) Abhishek Katiyar and Aman Khan gave me the key tip: Claude Code stores all your conversation history as text files on your computer. Open a new directory and give Claude Code a task. Here's how to watch compaction happening: 1. Go to your user's root directory 2. Press Command+Shift+Period (Mac) to show hidden folders 3. Navigate to ~/.claude/projects/ 4. Find your project folder and use Cursor/VSCode to open it (there's a reason) 5. Install the JSONL Gazelle plugin (open source, thank you Gabor Cselle!) 6. Open the most recent JSONL file - each row is a message in your conversation 7. Run the compact command in Claude Code with custom instructions 8. Watch what happens in the file What I learned: When you compact, Claude Code doesn't just summarize and delete everything. It creates a "compact boundary" in the conversation file, writes a summary of what happened before, but keeps the full original conversation (!!!!) The new thread can still retrieve any details from before compaction if needed. That is so damn cool. Why this matters: What you're getting in Claude Code is similar to what Anthropic ships in their developer SDK - so inspecting your daily tools is how you build real product intuition. The best way to understand AI systems is to open them up and look inside. Everything is text files.

Tal Raviv

57,910 views • 6 months ago

Using Claude Code as a general agent part 3: Getting Claude Code to Ask me clarifying questions... inline, by creating a Clarifying-Questions .md file What's in this file is how the agent should format clarifying questions before doing a specific task. Rather than coming up with ALL of the context I need to give to the agent, I state what I want at the top of the docuemnt then i tell claude to ask me clarifying questions IN THE DOCUMENT. The agent knows to format them as empty code blocks and all i need to do is fill them in then tell Claude Code to finish the task at hand. Kinda niche, but I use this for EVERYTHING now: 1. State what I want at top (Ad Scripts) 2. Type out relevant context that's in my head 3. Tell the agent to ask @ clarifying-questions 4. Quickly answer them with wispr flow in the document 5. Tell Claude to continue This would be useful for nearly every workflow.

Using Claude Code as a general agent part 3: Getting Claude Code to Ask me clarifying questions... inline, by creating a Clarifying-Questions .md file What's in this file is how the agent should format clarifying questions before doing a specific task. Rather than coming up with ALL of the context I need to give to the agent, I state what I want at the top of the docuemnt then i tell claude to ask me clarifying questions IN THE DOCUMENT. The agent knows to format them as empty code blocks and all i need to do is fill them in then tell Claude Code to finish the task at hand. Kinda niche, but I use this for EVERYTHING now: 1. State what I want at top (Ad Scripts) 2. Type out relevant context that's in my head 3. Tell the agent to ask @ clarifying-questions 4. Quickly answer them with wispr flow in the document 5. Tell Claude to continue This would be useful for nearly every workflow.

Riley Brown

27,664 views • 7 months ago

Claude certified architect highest paid AI job Claude announced their course to become a "claude certified architect" and you can become one for free today. many people are offering courses for tens of thousands, but now Claude is offering it all for free on their website. I started diving into their material and its impressive how deep it goes it covers everything from: > Agentic Architecture & Orchestration (27%) > Tool Design & MCP Integration (18%) > Claude Code Configuration & Workflows (20%) > Prompt Engineering & Structured Output (20%) > Context Management & Reliability (15%) sign up here:

Claude certified architect highest paid AI job Claude announced their course to become a "claude certified architect" and you can become one for free today. many people are offering courses for tens of thousands, but now Claude is offering it all for free on their website. I started diving into their material and its impressive how deep it goes it covers everything from: > Agentic Architecture & Orchestration (27%) > Tool Design & MCP Integration (18%) > Claude Code Configuration & Workflows (20%) > Prompt Engineering & Structured Output (20%) > Context Management & Reliability (15%) sign up here:

Tim Haldorsson

330,127 views • 4 months ago

I now spend more time building and far less time on outreach emails. Every campaign includes: • Finding prospects • Learning about each company • Personalizing every email • Writing follow-ups • Tracking replies That workflow can add up quickly—long before I even hit “Send.” So I tried Sintra. After I shared a bit of context about my business, it started generating personalized outreach sequences, follow-up emails, and messages tailored to different prospects—all in one place. Now, instead of starting from scratch, I review the drafts, make quick edits when needed, and send. The biggest difference isn’t just that AI helps write emails. It’s that I get my time back to focus on building while Sintra handles the repetitive parts. If outreach is a key part of your work, this is worth trying. 🔗 Use code: ZAYVEN75 for 75% off for lifetime.

I now spend more time building and far less time on outreach emails. Every campaign includes: • Finding prospects • Learning about each company • Personalizing every email • Writing follow-ups • Tracking replies That workflow can add up quickly—long before I even hit “Send.” So I tried Sintra. After I shared a bit of context about my business, it started generating personalized outreach sequences, follow-up emails, and messages tailored to different prospects—all in one place. Now, instead of starting from scratch, I review the drafts, make quick edits when needed, and send. The biggest difference isn’t just that AI helps write emails. It’s that I get my time back to focus on building while Sintra handles the repetitive parts. If outreach is a key part of your work, this is worth trying. 🔗 Use code: ZAYVEN75 for 75% off for lifetime.

ZAYVEN KNOX

22,820 views • 14 days ago

I totally relate with Meghan here! By 40 weeks of pregnancy I was a menace & wanted the babies out! I did everything possible to ‘encourage’ them out and I mean EVERYTHING! If I told you all the antics I got up to it would make you blush 😂 #MeghanMarkle #HarryAndMeghan

I totally relate with Meghan here! By 40 weeks of pregnancy I was a menace & wanted the babies out! I did everything possible to ‘encourage’ them out and I mean EVERYTHING! If I told you all the antics I got up to it would make you blush 😂 #MeghanMarkle #HarryAndMeghan

Dr Shola Mos-Shogbamimu

260,969 views • 1 year ago

Today, we launched our official Claude connector at Quo (formerly OpenPhone). Now you can get insights from every customer conversation, send bulk messages, update contacts, and so much more - all directly inside Claude. I've been using this every day to understand how things are going for our inbound sales team, what's working with prospects, and where the coaching opportunities are. With the Claude connector, I create a scheduled task that scores all the transcripts from our Quo phone number against 2 of our sales rubrics and outputs a report with a summary, score, and any coaching opportunities. Let's be honest, no manager has time to listen to and score 50 calls a day. But Claude does. Now I just check Slack every morning, where Claude delivers all the context I need. Every Quo customer with Claude now has this capability (and so much more). Connecting the tools takes minutes, and then work that used to require an Engineering team is only a prompt away. Huge shout-out to everyone on our team for shipping this! I can't wait to see what you build with it.

Today, we launched our official Claude connector at Quo (formerly OpenPhone). Now you can get insights from every customer conversation, send bulk messages, update contacts, and so much more - all directly inside Claude. I've been using this every day to understand how things are going for our inbound sales team, what's working with prospects, and where the coaching opportunities are. With the Claude connector, I create a scheduled task that scores all the transcripts from our Quo phone number against 2 of our sales rubrics and outputs a report with a summary, score, and any coaching opportunities. Let's be honest, no manager has time to listen to and score 50 calls a day. But Claude does. Now I just check Slack every morning, where Claude delivers all the context I need. Every Quo customer with Claude now has this capability (and so much more). Connecting the tools takes minutes, and then work that used to require an Engineering team is only a prompt away. Huge shout-out to everyone on our team for shipping this! I can't wait to see what you build with it.

Daryna Kulya

32,381 views • 2 months ago

This is how real Polymarket empires are built. Everyone is still looking for that one Claude setup that will change their life. I keep thinking that vibe coding does nothing, and instead of wasting time, I build a real product. While you believe it’s impossible to trade arbitrage on five-minute markets, MANUAL CODE can do it. I have already connected data on: - Current Bitcoin price. - Live five-minute pool odds. - Real-time order book. In test mode, the results are already promising. I’m waiting for the next stages of validation, and then it can be launched as a bot with market-based calls. Should I publish this?

This is how real Polymarket empires are built. Everyone is still looking for that one Claude setup that will change their life. I keep thinking that vibe coding does nothing, and instead of wasting time, I build a real product. While you believe it’s impossible to trade arbitrage on five-minute markets, MANUAL CODE can do it. I have already connected data on: - Current Bitcoin price. - Live five-minute pool odds. - Real-time order book. In test mode, the results are already promising. I’m waiting for the next stages of validation, and then it can be launched as a bot with market-based calls. Should I publish this?

Raychi

42,128 views • 4 months ago

2min love the sound of each other's voices and seungmin wrote cinema with a specific sound of lee know's voice in mind 😭😭😭 🐶: to add a bit about "cinema", there's this voice of yours that i really like and that voice stay really loves it too. that old, kind of tickly natural sound you just let out. i wanted to bring that back after a long time. so i wrote a lot while thinking about that. even when asked 3racha for advice, i asked a lot about which direction to pull it 🐶: so it would feel like "does this song and you blend well?" after i finished the song, recorded everything and played it for 3racha they said "did you write this thinking about lee know?" 🐰: back then, when i first heard seungmin's voice, early on, i remember thinking, "nice tone". the way he does it calmly like that, i always really liked listening to it

2min love the sound of each other's voices and seungmin wrote cinema with a specific sound of lee know's voice in mind 😭😭😭 🐶: to add a bit about "cinema", there's this voice of yours that i really like and that voice stay really loves it too. that old, kind of tickly natural sound you just let out. i wanted to bring that back after a long time. so i wrote a lot while thinking about that. even when asked 3racha for advice, i asked a lot about which direction to pull it 🐶: so it would feel like "does this song and you blend well?" after i finished the song, recorded everything and played it for 3racha they said "did you write this thinking about lee know?" 🐰: back then, when i first heard seungmin's voice, early on, i remember thinking, "nice tone". the way he does it calmly like that, i always really liked listening to it

lena ✰

14,277 views • 6 months ago

Introducing Claude Code Hook - Context Timeline (Saving this to try later) Install with: npx claude-code-templates@latest --hook monitoring/context-timeline Managing the context window and the subagents running in Claude Code is hard to keep track of That's why I built this hook... It starts the moment you open a session and shows a timeline with the main agent's context window and how subagents start working in their own separate context Every subagent you have running will show up in real time This way you can manage the context and the subagents you run, and see everything in a much simpler way than in the console

Introducing Claude Code Hook - Context Timeline (Saving this to try later) Install with: npx claude-code-templates@latest --hook monitoring/context-timeline Managing the context window and the subagents running in Claude Code is hard to keep track of That's why I built this hook... It starts the moment you open a session and shows a timeline with the main agent's context window and how subagents start working in their own separate context Every subagent you have running will show up in real time This way you can manage the context and the subagents you run, and see everything in a much simpler way than in the console

Daniel San

51,690 views • 2 months ago