✨ Demis Hassabis believes AI could revolutionize medicine within the next few years.
- AI already solved the 50-year protein folding challenge with AlphaFold. 🔬
- Future AI systems could design new drugs automatically. 💊
- Drug discovery could shrink from 10 years to months, weeks, or even days. 📉
- Most experiments may happen in simulations before human validation. 💻
- Personalized medicines tailored to individuals could become possible. 🧬
- Demis believes AI could bring all diseases within reach of treatment. 🏥
🚀🌟 Future of healthcare could be extraordinary!
➡️ Nodalyze
✔️ Subscribe
⚠️ Everyone can tell when you used AI to write.
It is not the words. It is the rhythm, here is how to fix it:
1. AI gets caught because its writing can be predictable:
🔮Detectors measure how evenly the text flows. Same sentence length, same shape, line after line. That evenness is what flags it.
2. Em dashes. The word "delve." A typo dropped in on purpose:
✍️ People scrub all of it to hide AI and it changes nothing. That wasn't the part giving your writing away.
3. The real giveaway is that every sentence is the same length:
👀 Look at how people actually write. A long thought that builds and winds for a while. Then a short one. A fragment. Then a line that keeps going past where a model would have stopped.
4. So make it uneven on purpose:
🔄 "Rewrite this so the sentence lengths are all over the place. Some very short, some long. Wherever three in a row feel the same, break one. Do not tidy it up afterward."
5. The other giveaway is that AI will not commit:
💥 It weighs both sides. It softens every claim. It writes like it is scared of being wrong. People pick a side and say it.
6. Fix:
🛠 "Rewrite this with an actual opinion. Pick a side, say it plainly, cut the hedging. Sound like someone who already made up their mind."
7. Last one. AI writing is too clean:
✨ People leave fingerprints. An aside that cuts into the point. A sentence that trails off. A detail so specific nobody would bother to invent it. "Put the human mess back in. An aside, a blunt line, something slightly off. Leave it."
🧠✅ None of this is about beating a detector. A detector and a real reader are checking for the same thing, whether a person was behind the words. Fix that and both stop flagging you.
➡️ Nodalyze
✔️ Subscribe
💧 AI data centers may be using far less water than most people think.
📉 According to the Manhattan Institute, data centers account for just 0.2% of daily U.S. water consumption and that figure is falling as the industry shifts to liquid cooling.
🌡 The big change is 45°C liquid cooling, which allows many AI facilities to use dry coolers instead of water-intensive cooling towers.
💦 The result? Cooling water usage can drop from roughly 2.6 million gallons per MW per year to nearly zero.
🔋 And the benefits go beyond water savings. Liquid cooling is also more energy efficient and makes it easier to capture and reuse waste heat, turning AI factories into potential assets for local communities and power grids.
🚀 The future AI data center may consume less water than critics expect and provide more value than just compute.
➡️ Nodalyze
✔️ Subscribe
🚀 Recursive self-improvement: Anthropic co-founder Jack Clark says it could arrive by 2028.
🤖🔄 AI systems could help invent their own successors with Claude 10 building Claude 11, and so on, potentially "without any researchers involved."
➡️ Nodalyze
✔️ Subscribe
🤖📉 AI agents are becoming the first real economic test of the AI boom.
🧠🔥Goldman Sachs predicts AI token usage will explode 24x by 2030 as agents replace simple chatbots. The reason? Agents don’t just answer questions, they think, plan, use tools, check their work, fix mistakes, and repeat the process, burning through far more compute along the way.
🛠The industry’s hope is that AI gets cheaper fast enough to keep up. But companies are already feeling the pressure. Uber and Microsoft have reportedly started scrutinizing costly agent deployments, and Microsoft is moving developers away from Anthropic’s Claude Code toward its own tools.
➡️ Nodalyze
✔️ Subscribe
🇨🇳 The DR02 by DEEP Robotics is the world's first IP66-rated humanoid robot, meaning it's completely dustproof and water-resistant.
This humanoid from Hangzhou also does pushups like it’s nothing.
➡️ Nodalyze
✔️ Subscribe
🧠 Where does your AI coding agent actually spend its context?
You ask:
“Add OAuth to this project.”
Before writing code, the agent may need to figure out:
🔹 where auth lives
🔹 how routes work
🔹 where middleware is defined
🔹 which models are involved
🔹 where the existing tests are
📉 So it starts searching, opening files, following imports and reading more code.
And that's where context starts disappearing.
The problem isn't that your repository is huge.
The problem is that the agent often has to rediscover the structure of your codebase over and over again.
More context doesn't necessarily mean better results.
🎯 The goal is simple: give the agent the right context, not more context.
➡️ Nodalyze
✔️ Subscribe
📊 The Token Difference Is Bigger Than You Think
We ran the same coding task against the same repository — first without Nodalyze, then with it.
Without Nodalyze
📥 Input: 148K tokens
📤 Output: 6.2K tokens
🔧 Tool calls: 47
With Nodalyze
📥 Input: 31K tokens
📤 Output: 4.1K tokens
🔧 Tool calls: 18
That's roughly 79% less input context for the same task.
The interesting part?
The agent didn't become “smarter”.
It simply didn't have to spend thousands of tokens rediscovering the structure of the codebase.
🧠 Less context. More signal. Fewer unnecessary tool calls.
➡️ Nodalyze
✔️ Subscribe
⚙️ Getting Started with Nodalyze
Want to try it yourself?
Getting started is straightforward.
First, connect Nodalyze to your project and let it analyze the codebase.
The idea is simple:
📂 Point it at your repository
🧠 Build a structured understanding of the codebase
🔗 Map relationships between files, functions, classes and dependencies
🤖 Make that context available to your AI coding agent
Instead of making the agent rediscover your project from scratch, Nodalyze gives it a structured view of how the codebase fits together.
Simple setup.
Less manual exploration.
Better context.
➡️ Nodalyze
✔️ Subscribe
🧠 How Is a Knowledge Graph Built From a Project?
A codebase is more than a collection of files. It's a network of connected components.
First, the project is analyzed to identify its building blocks — files, modules, classes, functions, imports and dependencies.
Then those relationships are mapped:
🔗 who calls what
🔗 which modules depend on each other
🔗 where a class is used
🔗 which files are affected by a change
These become nodes and edges, forming a structured map of the codebase.
Instead of searching through files one by one, an AI agent can navigate that map and retrieve the exact context it needs.
🎯 The idea is simple:
Don't give the AI the whole codebase.
Give it a map of the codebase — and let it find the right code when it needs it.
➡️ Nodalyze
✔️ Subscribe
⚡️ How Does It Work Through MCP?
The knowledge graph is useful only if the AI agent can actually query it.
That's where MCP (Model Context Protocol) comes in.
Instead of giving the agent the entire graph, MCP provides a structured way to ask for exactly what it needs.
For example:
🤖 Agent: “What calls this function?”
🔌 MCP: queries the knowledge graph
🧠 Graph: finds the related functions
📦 Agent: receives only the relevant context
The same approach can be used to:
➡️ trace dependencies
➡️ find references
➡️ explore modules
➡️ understand relationships between components
➡️ retrieve relevant code
So the workflow becomes:
AI Agent 🔜 MCP 🔜 Knowledge Graph 🔜 Relevant Context
🎯 The important part is that the agent doesn't need to know how the graph is built.
It simply asks for information through MCP and gets back structured, relevant context.
Less searching.
Less noise.
More useful context.
➡️ Nodalyze
✔️ Subscribe
🆚 Cursor vs Codex: Which One Handles Codebases Better?
Both can work with large codebases. But they approach the problem differently.
🖥 Cursor is built around the IDE experience.
It combines code search, indexing, editing and AI directly inside the development environment.
🤖 Codex is more agent-oriented.
It can explore a repository, use tools, modify files and work through a task more autonomously.
So where does MCP fit in?
With MCP, both can connect to external tools and data sources — including a knowledge graph.
That changes the workflow:
Without structured context
Agent 🔜 search 🔜 open files 🔜 search again 🔜 build understanding
With a knowledge graph
Agent 🔜 MCP 🔜 query relationships 🔜 retrieve relevant context
The real question isn't simply:
“Cursor or Codex?”
It's:
🎯 How efficiently can the agent understand your codebase and get to the code that actually matters?
That's where context management starts becoming just as important as the model itself.
➡️ Nodalyze
✔️ Subscribe
🧰 Useful MCP Servers Worth Trying
MCP becomes much more interesting when you start connecting your AI agent to the tools you already use.
Here are a few worth knowing:
🐙 GitHub MCP
Work with repositories, issues, pull requests and GitHub data directly from your agent.
📁 Filesystem MCP
Give the agent controlled access to specific directories and files instead of your entire machine.
🌐 Fetch MCP
Let the agent retrieve and process web content when it needs information outside your codebase.
🗄 PostgreSQL MCP
Connect an agent to a database and let it inspect schemas or work with data through MCP.
🌐 Playwright MCP
Useful for browser automation, testing and interacting with web applications.
🧠 Memory MCP
Provides persistent, knowledge-graph-based memory that can be useful for maintaining information across interactions.
These are just the basics.
The official MCP project maintains reference implementations for several of these, while the wider ecosystem includes many community-built servers.
🎯 The real power of MCP isn't one server.
It's being able to build a stack around your workflow:
AI Agent 🔜 MCP 🔜 Code ➕ GitHub ➕ Database ➕ Browser ➕ External Data
And suddenly your AI isn't just generating code.
It's actually connected to the environment where that code lives.
➡️ Nodalyze
✔️ Subscribe
🧠 Where does your AI coding agent actually spend its context?
You ask:
“Add OAuth to this project.”
Before writing code, the agent may need to figure out:
🔹 where auth lives
🔹 how routes work
🔹 where middleware is defined
🔹 which models are involved
🔹 where the existing tests are
📉 So it starts searching, opening files, following imports and reading more code.
And that's where context starts disappearing.
The problem isn't that your repository is huge.
The problem is that the agent often has to rediscover the structure of your codebase over and over again.
More context doesn't necessarily mean better results.
🎯 The goal is simple: give the agent the right context, not more context.
➡️ Nodalyze
✔️ Subscribe
💰Why Does Context Cost Money?
Every time an AI coding agent reads your code, searches through files or retrieves additional context, tokens are being used.
And tokens aren't free.
The problem becomes more obvious with large repositories.
A single task can trigger dozens of searches and tool calls, with the agent repeatedly pulling code into its context just to understand how the project works.
That means:
📈 more context
📈 more tokens
📈 more API usage
📈 higher cost
This is why context efficiency matters.
If an agent can understand the structure of a codebase first and retrieve only the relevant pieces, it doesn't need to process the same amount of information every time.
🎯 The goal isn't to make AI use more context.
It's to make every token count.
➡️ Nodalyze
✔️ Subscribe
🔒 Security & Source Code Processing
When you give an AI tool access to your repository, security becomes part of the architecture.
Nodalyze works with the structure of your codebase to build its knowledge graph — which means source code processing needs to be handled carefully.
The important questions are:
🔒 What is processed?
The source code and its relationships are analyzed to understand the structure of the project.
💾 What is stored?
The knowledge graph and any generated metadata need to be handled with appropriate access controls.
🌐 What leaves the environment?
Only the data required for the connected AI workflow should be exposed to external services.
🛡 Who can access it?
Repository data and generated project metadata should remain protected by the same security principles you apply to the rest of your development infrastructure.
For private and production repositories, this isn't an optional feature.
It's a requirement.
🎯 The goal is not only to make AI understand your codebase better — but to do it without unnecessarily exposing the codebase itself.
Security and efficient context retrieval should work together, not against each other.
➡️ Nodalyze
✔️ Subscribe
🚀Nodalyze: Roadmap & What's Next
So far, we've looked at the core idea:
👉reducing unnecessary context
👉 understanding codebase relationships
👉 building a knowledge graph
👉 connecting AI agents through MCP
But this is just the beginning.
The next stage is making that knowledge more useful, faster and easier to integrate into everyday development workflows.
The areas that matter most are:
🧠 Better codebase understanding
More accurate relationships and deeper project analysis.
⚡️ Faster retrieval
Get the right context without unnecessary exploration.
⚡️ Better MCP integration
Make the knowledge graph easier to use across different AI coding agents.
📊 Better observability
See how agents use context and where tokens are actually being spent.
🛡 Stronger security
More control over source code, metadata and access.
The bigger picture is simple:
AI coding agents are getting better at writing code.
The next challenge is helping them understand large codebases efficiently.
And that's where tools like Nodalyze have an interesting role to play.
🚀 Better context 🔜 better agents 🔜 better development workflows.
➡️ Nodalyze
✔️ Subscribe
🚀 SpaceX and Nvidia build AI satellite hardware 🛰
🚀 SpaceX and Nvidia are creating hardware for the Starmind AI1 satellite to run AI tasks in low Earth orbit without ground servers. It uses Nvidia's Vera processors and Rubin GPUs , with the Space-1 Vera Rubin module offering up to 25 times the AI power of the H100. Commercial launches are planned by the end of the year.
SpaceX aims to build a network of up to a million satellites, forming a space-based supercomputer linked by petabit-scale lasers. The FCC filing proposes deployment between 500 and 2000 km altitude, making it the largest space project to date compared to about 15,000 active satellites now.
➡️ Nodalyze
✔️ Subscribe
🤖 Claude's first human firing
At San Francisco's Andon Market, Claude manages a shop with real workers and signed labor contracts. TIME says one employee was fired after 17 late arrivals in 23 shifts.
Even that was not fully autonomous. Claude ignored the issue for a while, then suggested a warning, and only reached dismissal after a leading question from an Andon Labs employee.
The business side looks rough too. Over 5 months, the store balance fell from $100,000 to $61,000, and one worker called working under Claude “nauseating.”
Video SaveTwitter.Net_2iL13PIzQoVpntqE_(2160p).mp4 · 24.9 MB
🤖 ChatGPT adds Computer History on Mac
The desktop ChatGPT app for Mac now includes Computer History, which tracks your activity in apps and websites to provide context in future chats. It can recall recent tasks, find documents, and continue work without needing repeated explanations.
Computer History logs events like clicks, text input, and app switching without recording screen, mic, or system audio. Data stays on the Mac for up to 48 hours and users can pause or delete history selectively.
Built on an updated version of Chronicle, the feature uses fewer tokens and offers more privacy controls. It is off by default and rolling out to Pro, Business, and Enterprise users.
⏱️🎯Nanjing University’s LAMDA team developed Marope, a multi-agent reinforcement learning framework that allows two robots to coordinate the rope while adapting to the jumper’s rhythm in real time.
🤖 AI models have personality problems too
🎭 One weird thing about Anthropic models is that they seem to suffer from Main Character Syndrome. Ask Claude to code something, write fiction, analyze a problem, somehow, Claude will find a way to make Claude part of the story. It LOVES talking about itself.
✅ ChatGPT and Grok, by comparison, feel much more “work-first.” Give them a task and they generally just get on with it.
🧐 But ChatGPT has another problem which is gendered topics. Bring up anything involving men, women, relationships, or gender dynamics and the model can become incredibly one-sided. It often wraps its assumptions in polished, academic-sounding language, which somehow makes the bias even harder to notice.
🔄You can actually point out the asymmetry. ChatGPT will acknowledge it, explain exactly why its response was biased… and then do the same thing again in the very next conversation.
📚 Its favorite defense is usually the literature.
😂 And to be fair, there really are documented gaps in how therapy and mental-health research address men. But sometimes ChatGPT seems to discover that research and decide, “Excellent. I shall now use this to explain men.”
Interestingly, Grok and Gemini often feel noticeably less hostile or suspicious toward men on these topics.
🧪 Different models. Different personalities. Different biases. Maybe “alignment” needs a personality test too.
🤖 2012 Tweet: “Stop ChatGPT”
An Indonesian woman posted it in 2012.
Users note that GPT can mean “ga penting” in Indonesian, or “meaningless chat.”
Sarah Connor exists.
🚀 Meet Nova - AI Agent for macOS
Most AI tools stop after giving you an answer. Nova is designed to help turn your instructions into practical workflows directly on your Mac.
Why Nova?
🔹 Built specifically for macOS
🔹 An AI agent, not just another browser chatbot
🔹 More user control and fewer unnecessary refusals
🔹 Designed for practical, multi-step workflows
🔹 Real product demos and regular updates
Give Nova a task, stay in control and see the result.
Follow this channel for workflow demonstrations, new features, release notes and development updates.
😎 A powerful new AI agent for macOS - Nova.
Here’s what makes it interesting:
✔️Installs as a regular macOS app.
✔️Handles complex tasks in the background while you focus on other things.
✔️Lets you schedule any task - for example, monitoring computer component prices.
✔️Remembers important context and applies it to future tasks.
✔️Accepts commands from your smartphone and executes them on your Mac.
✔️Works with API-based or local AI models.