Nathan Hutchinson’s Post

I look forward to every conversation I have with Idan Gazit from GitHub Next. If you want direct access to a big picture thinker with a strong vision, don't miss this event at GitHub HQ in San Francisco on May 7. Pavan Tallapragada brings a view of real world platform engineering at the cutting edge. And Chris Romp is the first person I'd direct anyone to if they have questions about AI assisted software development. This event will take you beyond just AI coding.

𝗩𝗦 𝗖𝗼𝗱𝗲 𝗷𝘂𝘀𝘁 𝗴𝗼𝘁 𝗮 𝗾𝘂𝗶𝗲𝘁 𝗯𝘂𝘁 𝘀𝗶𝗴𝗻𝗶𝗳𝗶𝗰𝗮𝗻𝘁 𝗲𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝘂𝗽𝗴𝗿𝗮𝗱𝗲. VS Code 1.117 landed April 22 and the feature most enterprise orgs will care about is BYOK - Bring Your Own Key. If you are on Copilot Business or Enterprise, you can now connect your own model API keys directly into VS Code chat, with your admin setting the guardrails. ● 𝗕𝗬𝗢𝗞 - connect Anthropic, OpenAI, or other provider keys directly in VS Code; admins control availability via policy on GitHub.com; developers stay in the IDE they already use ● 𝗜𝗻𝗰𝗿𝗲𝗺𝗲𝗻𝘁𝗮𝗹 𝗿𝗲𝗻𝗱𝗲𝗿𝗶𝗻𝗴 - chat responses stream block-by-block now, which noticeably reduces perceived wait time on longer outputs ● 𝗕𝘂𝗶𝗹𝘁-𝗶𝗻 𝗖𝗼𝗽𝗶𝗹𝗼𝘁 𝗖𝗵𝗮𝘁 - from 1.116 (April 15): Copilot Chat is now a built-in VS Code extension; new users no longer need to install anything to get started BYOK removes the hard dependency on GitHub's model choices. You can route to whichever model performs best on your codebase without switching IDEs or toolchains. What model would you connect first - and has model lock-in been a real barrier for your team's VS Code Copilot adoption? #VSCode #GitHubCopilot #DeveloperTools

🤖 One of GitHub's most interesting updates this week is happening in the CLI: GitHub Copilot CLI now supports BYOK and local models⚡ What that means: ▶️ teams can connect their own model provider ▶️ they can also run fully local models ▶️ GitHub-hosted routing is no longer the only option ▶️ authentication is optional when using your own provider Why this matters: ✅ more flexibility in model choice ✅ more control over LLM spend ✅ better fit for restricted or air-gapped environments ✅ easier alignment with enterprise infrastructure choices This is a meaningful step for teams that want the Copilot CLI experience, but with more control over how and where inference happens. #GitHub #GitHubCopilot #CLI #AI #DeveloperTools

Ollama and GitHub Copilot. Another important union for privacy and power in the terminal.... The AI development landscape gains another chapter in its evolution. The announcement that Ollama now supports GitHub Copilot CLI reinforces an integration movement seeking the balance between cloud intelligence and local processing security. To date, no one holds an exclusive path to efficiency, but this combination of tools certainly opens new doors for developers. What this integration allows you to do now: - Repository Exploration. You can use Copilot CLI to map codebases and understand complex structures with the support of local processing. - Terminal Automation. This union allows for task planning based on GitHub tickets where AI assists in editing files and installing dependencies more fluidly. - Privacy and Control. By using Ollama as a backend option, developers gain another layer of choice regarding where their sensitive context should be processed. My personal analysis on this movement In my view, what we are witnessing is the consolidation of a hybrid model. Copilot's support for Ollama is another step acknowledging that the future of corporate software will not be centered on a single closed solution. My predictive analysis is that the terminal will remain the primary command center, now powered by agents that respect the security perimeter of each project. True productivity does not stem from a single tool, but from the ability to integrate the best available solutions into your workflow. #Ollama #GitHubCopilot #AI #OpenSource #CTO #SoftwareDevelopment #Privacy #TechTrends #Coding

When the GitHub Copilot SDK was released, it opened the door to entirely new ways of bringing AI into the software development lifecycle. One of my clients had a simple idea: How do we make the power of GitHub Copilot accessible to people who don’t live inside IDEs or CLIs? So I built WebSpec as a PoC/Accelerator (..and submitted it to the SDK Challenge)— a web application powered by the GitHub API and GitHub Copilot SDK that integrates with Copilot Spaces and WorkIQ to bring relevant organizational context into a simple, chat‑based experience. Now, teams can: Research solutions Create requirements Draft technical specifications …all without needing developer tooling. This solution was built using a Spec‑Driven Design approach. In the repository, you’ll find examples of Custom Agent, Instructions, Prompts & Skills. Plus you can checkout the generated specifications. Feel free to explore the repo to learn more: https://lnkd.in/eCh-FTG4 #github #copilot #sdk #workiq #learnitall

Remote control CLI sessions on web and mobile in public preview With the time savings that GitHub Microsoft Copilot unlocks, developers can focus more on the tasks they enjoy most—creative problem-solving, innovative projects, and experimentation. GitHub has introduced remote session capabilities for the GitHub Copilot CLI, which is currently in public preview. The new copilot --remote feature allows developers to monitor and interact with an active CLI session directly from web or mobile applications, providing real-time visibility and control while keeping actions synchronized across environments. Key capabilities include: • Sending instructions during execution and continuing workflows seamlessly • Reviewing and adjusting execution plans before implementation • Switching between plan, interactive, and autopilot modes • Managing permission requests within existing CLI policies • Responding to prompts generated during execution This update signifies a broader shift toward more flexible, agent-driven development workflows, enabling engineers to manage long-running tasks without being limited to a single device. It represents a meaningful step forward in enhancing developer productivity and workflow continuity. #GitHub #Copilot #DeveloperTools #AI #SoftwareEngineering

Stuck in a meeting but need code shipped? Send an issue to GitHub Copilot Coding Agent directly from Microsoft Teams. By the time your meeting ends, it has a PR ready. → Mention @github in Teams with your prompt → Copilot Coding Agent picks it up → Writes the code, opens a PR → You review when you are free This is what the agentic workflow actually looks like in practice. Not a demo. Not a concept. A real Teams → GitHub → PR pipeline. Meetings that ship code. What a time to be alive. #GitHubCopilot #MicrosoftTeams #CodingAgent #AgenticAI #DeveloperProductivity #AI

GitHub Copilot CLI… without a GitHub Copilot subscription? With recent restrictions around GitHub Copilot's new personal subscriptions sign-up, I wanted to see if I could still use the CLI experience with open models. Turns out—you can. I set up Copilot CLI using Open LLMs (BYOK approach), and it opens up a lot of possibilities: • Run models locally (depending on your hardware) • Use hosted models via OpenRouter • Control cost and performance • Customize how AI interacts with your workflows I also went deeper into: Context window management in CLI Model selection (weights, quantization, parameters) LM Studio configuration for running open source models And for a practical demo 👇 I built a skill that converts a wireframe into HTML/React code using the CLI. This setup is surprisingly powerful if you want flexibility beyond standard GH Copilot usage. 🎥 Watch here: https://lnkd.in/eHGM3b97 Let me know in the comments if you would like a copy of the wireframe to HTML code skill. #AI #Copilot #LLM #OpenSource #DeveloperTools #SoftwareEngineering

GitHub Copilot gets called an autocomplete tool. That undersells it significantly. Yes, it suggests code inline as you type. But the more interesting capability is what it does at the workspace level — it can read your entire codebase, answer questions about it, explain unfamiliar code, and act as an agent that spans multiple files. The numbers back up the adoption: 1.8M+ paid subscribers, integrations across every major IDE, and deep GitHub ecosystem access that no third-party coding agent can replicate natively. If you're evaluating AI coding tools, GitHub Copilot is the baseline everything else gets compared to. Full profile and alternatives → https://lnkd.in/ewkPwZeA

🧠 𝗔𝗴𝗲𝗻𝘁𝘀 𝗮𝗿𝗲 𝗯𝗿𝗲𝗮𝗸𝗶𝗻𝗴 𝗳𝗹𝗮𝘁-𝗿𝗮𝘁𝗲 𝗽𝗿𝗶𝗰𝗶𝗻𝗴 — 𝗮𝗻𝗱 𝗚𝗶𝘁𝗛𝘂𝗯 𝗷𝘂𝘀𝘁 𝘀𝗮𝗶𝗱 𝗶𝘁 𝗼𝘂𝘁 𝗹𝗼𝘂𝗱 GitHub announced major changes to Copilot Individual plans last week. New sign-ups paused, tighter usage limits, Opus 4.7 restricted to Pro+. But the why is the part worth reading 👇 Quote from the announcement: "It's now common for a handful of requests to incur costs that exceed the plan price." That's the AI-coding economy in one sentence. What it means in practice: ✅ Long-running, parallelized agent sessions = real infrastructure cost ✅ A single /fleet run can burn more tokens than a week of autocomplete ✅ Token-based weekly caps are the new reality, on top of premium request quotas ✅ Model access is becoming a tier differentiator (Opus → Pro+ only) 🔗 https://lnkd.in/dsKcnirg #GitHubCopilot #AIAgents #VibeCoding #SoftwareEngineering #DevTools #MVPBuzz