Anthropic just released another incredible coding model — and the speed is unbelievable.

The new Claude Haiku 4.5 gives you all the intelligence of Sonnet 4.5 at only a fraction of the cost.

It’s leaner, it’s faster, it’s cheaper — and it even comes with a brand new feature never seen before in this line of models.

Haiku 4.5 is built for pure speed. Speed is what this is all about.

This model is perfect for anything needing a real-time response — chat assistants, customer service bots, or coding copilots.

IDEs like Cursor have already added support for it — to make your development even smoother with quicker responses.

Being significantly faster than Sonnet 4.5 definitely makes for a much snappier user experience. Near-premium performance for much less.

This is a model that can be deployed at massive scale — making powerful AI more accessible for anyone looking to manage costs — even powering free-tier user experiences.

And the new Extended Thinking ability makes it even better.

With Extended Thinking you can boost complex reasoning and get even better accuracy with your coding — thinking harder to handle multi-step problems that would have previously required a slower, larger model like Sonnet.

Haiku 4.5 even supports Computer Use — which means it can autonomously use your PC to carry out complex tasks.

Manage files, fill out forms, handle spreadsheets and email… the possibilities are endless.

Haiku 4.5 processes both text and images and supports large context windows — up to ~200k tokens with generous output — making it suitable for manage full codebases and image-assisted tasks.

It also scores very well in industry benchmarks like SWE-bench.

Its speed also makes it ideal for multi-agent systems, where a larger model like Sonnet 4.5 could handle the overall plan, and multiple Haiku agents execute the parallel subtasks — like code refactoring or document analysis.

Claude Haiku 4.5 is a major step forward in making high-performance AI more affordable and deployable.

It’s not just a smaller sibling to Sonnet — it’s a purpose-built engine for real-time intelligence.

With new features like Extended Thinking and Computer Use it bridges the gap between speed and sophistication — giving you high-power state-of-the-art intelligence at scale.

Gemini CLI just got even more incredible with this massive new upgrade from Google.

Now with Extensions, you can connect Gemini CLI straight to your favorite tools— Nano Banana, BigQuery, Genkit, anything —and use them without ever leaving the terminal.

Look how we easily wrote several Postman tests for our APIs right from Gemini CLI — nothing but a single prompt:

They were all there for us to run at any time:

Extensions are like plugins for Gemini CLI — built with MCP servers.

With Gemini CLI Extensions you can install integrations with a single command and manage them easily — and even build your own.

They teach the AI how to work with specific tools, APIs, or services, so it can generate code, run commands, and answer questions with context.

And there are boatloads to choose from — with many more on the way — both official and from the community.

For example you could install a BigQuery extension that lets Gemini write SQL for your datasets, or a Genkit extension that scaffolds app logic and helps debug workflows.

Installing an extension is easy:

You can also list, enable, or disable extensions like this:

Each extension includes a simple manifest file (gemini-extension.json) that tells Gemini how to use the tools it exposes. If something’s wrong with the manifest, Gemini CLI will flag the issue immediately so you can fix it.

What else is new

Here’s what shipped with Extensions:

Official announcement & early partners

Gemini CLI now works with tools from Dynatrace, Elastic, Shopify, and others.

Public extensions gallery

You can now browse and install featured extensions from geminicli.com

Google Cloud integrations

Tools for BigQuery and Data Cloud let you query datasets, generate SQL, and run analysis—all in natural language.

Genkit support

The official Genkit extension lets Gemini CLI scaffold features, debug logic, and interact with Genkit’s dev flow.

Image workflows

Extensions like Nano Banana let Gemini generate and manipulate images, directly in the terminal.

It’s huge

This is a big leap forward for the Gemini CLI.

• Fewer tabs, less context-switching: Everything happens in the terminal. You don’t need to search for docs or flip through dashboards.

• More relevant help: Because extensions include tool definitions and context, Gemini gives you answers that actually fit your stack.

• Custom workflows: You can install multiple extensions—security, observability, analytics—and use them together in the same prompt.

Amazing extensions to try

If you’re not sure where to start, these are some really useful extensions to install:

1. Security

Scan your codebase for vulnerabilities.

gemini extensions install https://github.com/gemini-cli-extensions/security

Then prompt Gemini:

“Scan my current repo for security issues and suggest fixes.”

2. BigQuery / Data Cloud

Analyze data or generate SQL just by describing what you want.

3. Genkit

Great if you’re building apps with Genkit. Gemini can guide and assist using the built-in tools.

4. Nano Banana

Generate images on the fly with Gemini Flash.

gemini extensions install https://github.com/gemini-cli-extensions/nanobanana

Try: “Create a 1024px blog header showing a data pipeline.”

Build your own?

You can create your own extension in less than 15 minutes.

Here’s the basic process:

1. Create a folder with a gemini-extension.json file describing your tools and prompts.
2. Define your tools—APIs, scripts, commands—with clear inputs and outputs.
3. Install locally with gemini extensions install . and debug if needed.
4. Share it by pushing to GitHub so others can install it with a URL.

The new docs and community examples make the process much easier. You can even submit your extension to the official gallery.

Gemini CLI is growing beyond a simple prompt—it’s becoming a hub for development workflows. With extensions, you can bring your own tools, reduce friction, and keep everything in one streamlined loop.

Whether you’re querying a dataset, debugging an app, scanning for vulnerabilities, or creating visuals, Gemini CLI extensions give you a flexible way to do more—with less effort.

If you want help picking extensions for your specific stack (like testing, analytics, or DevOps), I can put together a custom starter kit for you.

This mind-blowing open-source tool is saving developers boatloads of hours every week — which is why so many people have been talking about it lately.

This is state-of-the-art automation – effortlessly connect several different tools, APIs and even AI models into one powerful workflow.

It’s like Zapier or Make but far better — far far more flexibility and control — they actually built this with coders in mind.

It’s open, powerful, and gives you the option to self-host it on your own server if you want to keep everything private.

How it all works

n8n is built around a simple concept: workflows made of nodes.

Each workflow starts with a trigger (for example, a new form submission, an incoming webhook, or a scheduled event).

From there you chain together nodes — each representing an action, such as calling an API, transforming data, sending an email, or running custom JavaScript.

Here’s what a typical n8n workflow might look like:

• When a new order comes in via your website →
• Fetch customer data from your database →
• Ask an AI model to generate a summary →
• Post that summary into Slack.

Everything flows visually, step by step, so you can see exactly how data moves through the process.

Hundreds of integrations and counting

n8n already connects to 500+ popular apps — from Google Sheets and Notion to GitHub, Slack, Airtable, and more.

And even if there’s no official integration, you can always use the HTTP Request or Database nodes to connect to almost any API or data source.

This flexibility means you can automate things across your entire stack — not just between mainstream SaaS tools.

The AI effect

Unlike traditional automation platforms n8n is built to work seamlessly with AI.

You can drop in nodes for OpenAI, Anthropic, Hugging Face, or custom LLM APIs to create “AI agents” that think and act inside your workflows.

For example:

• When a new customer email arrives →
• Use GPT-5 to summarize and classify the message →
• Log it in your CRM and auto-draft a response →
• Send it for review or auto-approve if it meets your rules.

Because everything is in one workflow, you can easily control, validate, or retry AI outputs — no black boxes.

Run it anywhere

One of n8n’s biggest advantages is freedom.

You can use n8n Cloud, their hosted version (no setup required), or run it yourself using Docker, Kubernetes, or a plain server. Self-hosting gives you full control over your data and customization, while the cloud version is perfect if you just want to start building right away.

Interesting licensing

n8n isn’t fully open-source — it’s released under a Sustainable Use License (SUL).

That means you can use, modify, and self-host it freely, but you can’t resell it as your own competing service without a commercial license.

It’s a “fair-code” model — a middle ground between open source and commercial software — meant to keep the company sustainable while still empowering developers.

What you can build with n8n

Here are some real-world examples of what people use n8n for:

• Data automation: pull from a database, clean it, and send reports to Google Sheets.
• Business operations: connect Stripe, Notion, and Slack for smoother workflows.
• AI content pipelines: generate summaries, captions, or emails automatically — with review steps built in.
• System monitoring: get instant Slack alerts when errors occur or thresholds are crossed.

Basically, if a task involves data moving between tools — you can probably automate it with n8n.

Why developers love it

What makes n8n special is that it strikes a balance between no-code and code. You can visually design 90% of your logic but still drop in JavaScript functions anywhere for fine control.

It also has advanced features like:

• Loops, branches, and conditional logic
• Sub-workflows and error handling
• Built-in variables and data mapping
• Retry and wait mechanisms

This makes it powerful enough for production-grade systems — not just hobby projects.

Who it’s for

n8n is perfect if you:

• Want more control than Zapier or Make can give
• Need to run automation on your own infrastructure
• Want to mix AI and traditional automation in one place
• Enjoy building scalable workflows visually

If you just need one-step “when-this-then-that” automations, simpler tools might do. But when your processes get complex, n8n gives you the structure and power to handle it.

n8n is quietly becoming one of the most flexible and developer-friendly automation tools out there. It’s not just about connecting apps — it’s about building systems that think and act for you.

Whether you’re a solo maker, a dev team, or a company looking to streamline your operations, n8n gives you the canvas to turn your workflows into living, automated systems.

SaaS is everywhere. Subscription costs add up fast. Open-source offers a powerful solution. These tools provide control and savings. Let’s explore 20 options to cut your SaaS expenses.

1. Supabase

It’s an open-source Firebase alternative. Build and scale easily.

Key Features:

• Managed PostgreSQL Database: Reliable and less operational hassle.
• Realtime Database: Live data for interactive apps.
• Authentication and Authorization: Secure user management built-in.
• Auto-generated APIs: Faster development from your database.

2. PocketBase

A lightweight, all-in-one backend. Setup is incredibly simple.

Key Features:

• Single Binary Deployment: Easy to deploy.
• Built-in SQLite Database: Fast and no extra install.
• Realtime Subscriptions: Reactive UIs are simple.
• Admin Dashboard: Manage data visually.

3. Dokku

Your own mini-Heroku. Deploy apps easily on your servers.

Key Features:

• Git-Based Deployments: Deploy with a Git push.
• Plugin Ecosystem: Extend functionality easily.
• Docker-Powered: Consistent environments.
• Scalability: Scale your apps horizontally.

4. Airbyte

Open-source data integration. Move data between many sources.

Key Features:

• Extensive Connector Library: Connect to hundreds of sources.
• User-Friendly UI: Easy pipeline configuration.
• Customizable Connectors: Build your own if needed.
• ELT Support: Simple to complex data movement.

5. Appwrite

A self-hosted backend-as-a-service. Build scalable apps with ease.

Key Features:

• Database and Storage: Secure data and file management.
• Authentication and Authorization: Robust user access control.
• Serverless Functions: Run backend code without servers.
• Realtime Capabilities: Build interactive features.

6. Ory Kratos

Open-source identity management. Security and developer focus.

Key Features:

• Multi-Factor Authentication (MFA): Enhanced security for users.
• Passwordless Authentication: Modern login options.
• Identity Federation: Integrate with other identity systems.
• Flexible User Schemas: Customize user profiles.

7. Plane

Open-source project management. Clarity and team collaboration.

Key Features:

• Issue Tracking: Manage tasks and bugs effectively.
• Project Planning: Visualize timelines and sprints.
• Collaboration Features: Easy team communication.
• Customizable Workflows: Adapt to your processes.

8. Coolify

A self-hosted PaaS alternative. Simple deployment of web apps.

Key Features:

• Simplified Deployment: Deploy with a few clicks.
• Automatic SSL Certificates: Free SSL via Let’s Encrypt.
• Resource Management: Monitor and scale resources.
• Support for Multiple Application Types: Versatile deployment.

9. n8n

Free, open-source workflow automation. Connect apps visually.

Key Features:

• Node-Based Visual Editor: Design workflows easily.
• Extensive Integration Library: Connect to many services.
• Customizable Nodes: Integrate with anything.
• Self-Hostable: Full data control.

10. LLMWare

Build LLM-powered applications. Open-source tools and frameworks.

Key Features:

• Prompt Management: Organize and test prompts.
• Data Ingestion and Indexing: Prepare data for LLMs.
• Retrieval Augmented Generation (RAG): Ground LLM responses.
• Deployment Options: Flexible deployment choices.

11. LangchainJS

JavaScript framework for language models. Build complex applications.

Key Features:

• Modular Architecture: Use individual components.
• Integration with Multiple LLMs: Supports various providers.
• Pre-built Chains and Agents: Ready-to-use logic.
• Flexibility and Extensibility: Customize the framework.

12. Trieve

Open-source vector database. Efficient semantic search.

Key Features:

• Efficient Vector Storage and Retrieval: Fast similarity search.
• Multiple Distance Metrics: Optimize search accuracy.
• Metadata Filtering: Refine search results.
• Scalability: Handles large datasets.

13. Affine

Open-source knowledge base and project tool. Notion and Jira combined.

Key Features:

• Block-Based Editor: Flexible content creation.
• Database Functionality: Structured information management.
• Project Management Features: Task and progress tracking.
• Interlinking and Backlinks: Connect your knowledge.

14. Hanko

Open-source passwordless authentication. Secure and user-friendly.

Key Features:

• Passwordless Authentication: Secure logins without passwords.
• WebAuthn Support: Industry-standard security.
• User Management: Easy account and key management.
• Developer-Friendly APIs: Simple integration.

15. Taubyte

Open-source edge computing platform. Run apps closer to users.

Key Features:

• Decentralized Deployment: Deploy across edge nodes.
• Serverless Functions at the Edge: Low-latency execution.
• Resource Optimization: Efficient resource use.
• Scalability and Resilience: Robust and scalable apps.

16. Plausible

Lightweight, privacy-friendly web analytics. An alternative to Google.

Key Features:

• Simple and Clean Interface: Easy-to-understand metrics.
• Privacy-Focused: No cookies, no personal tracking.
• Lightweight and Fast: Minimal impact on site speed.
• Self-Hostable: Own your data.

17. Flipt

Open-source feature flags and experimentation. Safe feature rollouts.

Key Features:

• Feature Flag Management: Control feature visibility.
• A/B Testing: Run controlled experiments.
• Gradual Rollouts: Release features slowly.
• User Targeting: Target specific user groups.

18. PostHog

Open-source product analytics. Understand user behavior.

Key Features:

• Event Tracking: Capture user interactions.
• Session Recording: See how users behave.
• Feature Flags: Integrated feature control.
• A/B Testing: Experiment and analyze.

19. Logto

Open-source authentication and authorization. Modern app security.

Key Features:

• Flexible Authentication Methods: Various login options.
• Fine-Grained Authorization: Granular access control.
• User Management: Easy user and permission management.
• Developer-Friendly SDKs: Simple integration.

20. NocoDB

Open-source no-code platform. Turn databases into spreadsheets.

Key Features:

• Spreadsheet-like Interface: Familiar data interaction.
• API Generation: Automatic REST and GraphQL APIs.
• Form Builders: Create custom data entry forms.
• Collaboration Features: Teamwork on data and apps.

The open-source world offers great SaaS alternatives. You can cut costs and gain control. Explore these tools and free yourself from high SaaS bills. Take charge of your software stack.

Claude Sonnet 4.5 totally stole the show so nobody is paying attention to this incredible tool that came with it.

This underrated tool could actually end up completely transforming the app ecosystem forever.

Meet Imagine with Claude — a revolutionary new tool for building software — apps that are ALIVE.

Apps that write their own code — are you with me??

First of all describe whatever you want and Claude builds it on the fly — there is no underlying static codebase anywhere.

And from then on — the software generates itself.

There is no compilation or building of pre-written code — the app generates more itself in real-time as you interact with it.

Instantly turn your wild ideas into working prototypes and beyond — Tweak in any way you want and shape the end result directly.

The key distinction here is that nothing is prewritten or predetermined.

When you click a button or enter text in the environment, Claude interprets your action and generates the necessary software components to adapt and respond instantly.

It’s software that evolves based on your in-the-moment needs, which is a significant departure from static, pre-packaged applications.

This is kind of magic that’s now possible thanks to this incredible new Claude Sonnet 4.5 model.

This new version of Claude is tuned for long, multi-step reasoning and tool use.

It literally coded non-stop for 30+ hours, which is just absolutely wild.

It’s blowing every competitor out of the water when it comes to Computer Use — autonomously performing tasks with your PC.

“Imagine with Claude” is a showcase for those abilities.

It’s a short-term experiment but its implications are huge for devs and designers and everyone else.

It points to a future of disposable, adaptive software.

Imagine needing a very specific, one-off tool for a task—instead of hunting for a pre-made solution or coding it yourself, the AI could instantly assemble a custom application that functions exactly how you need it to, right when you need it.

It essentially collapses the gap between idea and working prototype to mere seconds.

Product designers could use it to create complex, interactive user interfaces instantly, allowing for faster feedback and iteration.

Users could generate specialized software to manage personal data, analyze complex information, or automate niche tasks without ever touching a line of code.

It’ll be really exciting to see just how much of an impact this has.

GitHub finally released their Copilot coding agent to the world and it’s been completely insane.

This is in a completely different league from agentic tools like Windsurf or Cursor.

Forget autocomplete and forget vibe coding — this is a full-blown genius teammate.

This is an AI companion you can delegate real development work to — and it comes back with a pull request for you to review.

It’s actually collaborating with you like a real human.

No need for endless prompting — just assign it massive tasks — like a entire GitHub issue.

And that’s that — you don’t have to guide or micromanage it in anyway.

The agent:

• Spins up an isolated GitHub Actions environment
• Clones your repo
• Builds and tests your code
• Opens a draft pull request with its changes.

Make comments on the PR and it will instantly make any needed changes.

It’s built to handle real tasks — not just making edits here and there:

Fixing major bugs, implementing features, improving test coverage, updating documentation, and so much.not just making edits here and there.

But the biggest selling point here the asynchronous delegation.

You’re no longer chained to your IDE while an AI tool generates code. You can:

• Offload routine work and keep coding on something else.
• Get a PR-first workflow that matches how your team already ships software.
• Run tasks in a clean CI-like environment, avoiding “works on my machine” issues.

Regular coding agents are amazing — but they live inside your editor. You’re chatting with them right there in your workspace.

They watch what you’re doing, keep track of your Problems panel, your edits, your clipboard — and they act instantly on your files. It’s like having a very attentive pair programmer who’s always sitting next to you.

But this Copilot agent doesn’t sit inside your IDE at all.

You hand it a task and it disappears into the cloud, does the work, and comes back later with all the results.

Instead of direct file edits you get a packaged, ready-to-review PR.

• Copilot Coding Agent is best for: Bug fixes with clear repro steps, test coverage boosts, doc updates, dependency bumps, or any feature slice you want to run in the background and review later.
• IDE Agents: Rapid prototyping, design-heavy changes, multi-file refactors, or anything where you want immediate feedback and full control.

Real examples:

• Refactor an API call across dozens of files — it branches, updates, tests, and PRs.
• Add a new endpoint with proper routing and unit tests.
• Migrate a dependency with code updates across the repo.

The new Copilot coding agent makes async, repo-level development feel seamless.

If Windsurf and Cursor are about collaborating with AI inside your IDE, Copilot’s agent is about giving your AI its own seat at the table — one that files branches and PRs just like a real developer.

It’s an entirely new way to build software — and it’s here now.

This will absolutely transform the MCP ecosystem forever.

Microsoft just released a new feature that makes creating MCP servers easier than ever before.

Now with Logic Apps as MCP servers — you can easily extend any AI agent with extra data and context without writing even a single line of code.

String together thousands of tools and give any LLM access to all the data flowing through them.

From databases and APIs to Slack, GitHub, Salesforce, you name it. Thousands of connectors are already there.

Now you can plug a whole new world of prebuilt integrations straight into your AI workflows.

Until now, hooking an LLM or agent up to a real-world workflow was painful. You had to write API clients, handle OAuth tokens, orchestrate multiple steps… it was a lot.

With Logic Apps as MCP servers, all that heavy lifting is already done. Your agent can call one MCP tool, and under the hood Logic Apps will ping APIs, transform data, or trigger notifications across your services.

You can wire up a Logic App that posts to social media, updates a database, or sends you alerts, and then call it from your AI app. No new server, no SDK headaches.

Microsoft’s MCP implementation even supports streaming HTTP and (with some setup) Server-Sent Events. That means your agent can get partial results in real time as Logic Apps run their workflows — great for progress updates or long-running tasks.

Because it’s running inside Azure, you get enterprise-grade authentication, networking, and monitoring. Even if you’re small now, this matters when you scale or if you’re dealing with sensitive data.

What can you do right now

• Build a Logic App that starts with an HTTP trigger and ends with a Response action.
• Flip on the MCP endpoint option in your Logic App’s settings.
• Register your MCP server in Azure API Center so agents can discover it.
• Point your AI agent to your new MCP endpoint and start calling it like any other tool.

Boom — your no-code workflow is now an AI-callable tool.

Some ideas to get you started

• Personal Dashboard: Pull data from weather, GitHub, and your to-do list, and serve it to your AI bot in one call.
• Social Blast: Draft tweets or LinkedIn posts with AI, then call a Logic App MCP server to publish them automatically.
• File Pipeline: Resize images, upload to storage, and notify a channel — all triggered by a single MCP call.
• Notifications & Alerts: Have your AI assistant call a Logic App to send you Slack, Teams, or SMS updates.

The bigger picture

This move is a major milestone because it connects two worlds:

• The agent/tooling world (MCP, AI assistants, LLMs)
• The workflow/integration world (Logic Apps, connectors, automations)

Until now these worlds were separate. Now they’re basically plug-and-play.

Microsoft is betting that MCP will be the standard for AI agents the way HTTP became the standard for the web.

By making Logic Apps MCP-native, they’re giving you a shortcut to a huge ecosystem of integrations and enterprise workflows.

You’re missing out big time if you’re still ignoring this incredible tool.

There’s so much it can do for you — but many devs aren’t even using it anywhere close to it’s fullest potential.

If you’ve ever wished your terminal could think with you — plan, code, search, even interact with GitHub — that’s exactly what Gemini CLI does.

It’s Google’s command-line tool that brings Gemini right into your shell.

You type, it acts. You ask, it plans. And it works with all your favorite tools — including being powered with the same tech behind the incredible Gemini Code Assist:

It’s ChatGPT for your command line — but with more power under the hood.

A massive selling point has been the MCP servers — acting as overpowered plugins for Gemini CLI.

Hook it up to GitHub, a database, or your own API, and suddenly you’re talking to your tools in plain English. Want to open an issue, query a database, or run a script? Just ask.

How to get started fast

Just:

You’ll be asked to sign in with your Google account the first time. Pick a theme, authenticate:

And you’re in:

Talking to Gemini CLI

There are two ways to use it:

• Interactive mode — just run gemini and chat away like you’re in a terminal-native chat app.
• Non-interactive mode — pass your prompt as a flag, like gemini -p “Write a Python script to…”. Perfect for scripts or quick tasks.

Either way, Gemini CLI can do more than just text. It can:

• Read and write files in your current directory.
• Search the web.
• Run shell commands (with your permission).

The secret sauce

Here’s where it gets exciting. MCP (Model Context Protocol) servers are like power-ups. Add one for GitHub and you can:

• Clone a repo.
• Create or comment on issues.
• Push changes.

Add one for your database or your docs, and you can query data, summarize PDFs, or pull in reference material without leaving the CLI.

All you do is configure the server in your settings.json file. Gemini CLI then discovers the tools and lets you use them in natural language.

Give Gemini a memory with GEMINI.md

Create a GEMINI.md in your project and drop in your project’s “personality.” It can be as simple as:

Always respond in Markdown.
Plan before coding.
Use React and Tailwind for UI.
Use Yarn for NPM package installs

Next time you run Gemini CLI, it will follow those rules automatically. You can check what memory it’s using with /memory show.

Slash commands = Instant prompts

If you do the same thing a lot — like planning features or explaining code — you can create a custom slash command.

Make a small TOML file in .gemini/commands/ like this:

description = “Generate a plan for a new feature”
prompt = “Create a step-by-step plan for {{args}}”

Then in Gemini CLI just type:

/plan user authentication system

And boom — instant output.

Real-world examples

Here’s how people actually use Gemini CLI:

• Code with context — ask it to plan, generate, or explain your codebase.
• Automate file ops — have it sort your downloads, summarize PDFs, or extract data.
• Work with GitHub — open issues, review PRs, push updates via natural language.
• Query your data — connect a database MCP server and ask questions like a human.

Safety first

Gemini CLI can run shell commands and write files, but it always asks first. You can allow once, always, or deny. It’s like having a careful assistant who double-checks before doing anything risky.

Gemini CLI isn’t just another AI interface. It’s a workbench where you blend AI with your existing workflows. Instead of hopping between browser tabs, APIs, and terminals, you get one cohesive space where you talk and it acts.

Once you add MCP servers, GEMINI.md context, and slash commands, it starts to feel less like a tool and more like a teammate who lives in your terminal.