LIKHITH ADITHYA’s Post

Claude Code users have had this for a while — now GitHub Copilot CLI is catching up 🤖 GitHub just launched remote session support for Copilot CLI in public preview. That means you can kick off a long agentic task on your machine, walk away, and monitor or steer it from your phone. In this week's Two Minute Tuesday I cover: ✅ How copilot remote CLI works ✅ What you can do from a remote session ✅ What admins need to enable first (Copilot Business/Enterprise) ✅ Tips to get started today This is the direction all AI tooling is heading. Worth getting familiar with it now. 🎥 Watch here: https://lnkd.in/eZNUKTX5 Are you in the GitHub Copilot camp, the Claude Code camp, or something else entirely? Drop it in the comments 👇 #GitHubCopilot #CopilotCLI #ClaudeCode #AITools #ITPro #TwoMinuteTuesdays

Environment strategy is the foundation of a mature Power Platform practice — and it's the conversation most teams skip until something breaks in production. Early in my career I watched a well-intentioned maker overwrite a production Power App while trying to fix a bug. No version history. No rollback. Just an apology email to 200 users. That's when I started taking ALM seriously. A baseline environment strategy that actually works in practice: — Dev → where makers build and experiment freely — Test/UAT → where stakeholders validate before release — Production → locked down, changes only via managed solutions Solutions should be publisher-prefixed, unmanaged in Dev, managed in Test and Prod. Pipelines in Power Platform (now generally available) can automate the promotion between environments. This isn't over-engineering. It's the difference between a platform your organization trusts and one they're afraid to touch. If you're building anything beyond a personal productivity app, your environment strategy conversation should happen on day one. #PowerPlatform #ALM #PowerApps #MicrosoftPlatform #EnterpriseArchitecture

Contelligence is a great example of taking a powerful capability and making it genuinely useful beyond a narrow audience. Turning everyday operational tasks into something automated and structured without adding complexity. What I appreciate most is the practical thinking behind it: clear use cases, real problems being solved, and a focus on usability from day one. Looking forward to seeing how this evolves! Stephen Helwig

A busy week for GitHub: code quality, Copilot upgrades, and CLI sessions go remote. - Code scanning alerts can now be linked directly to GitHub Issues, making it easier to track and remediate vulnerabilities within normal planning flows. This preview also brings visual tracking status and new filters to help teams focus on untracked alerts. https://lnkd.in/diaF7bX2 - Model selection for Claude and Codex agents is live, so you can pick Anthropic or OpenAI models for coding tasks just like with Copilot. Admins control access, and support spans current Claude Sonnet, Opus, and GPT-5.x variants. https://lnkd.in/dGRdpphH - GitHub Code Quality gets big usability improvements in public preview. Findings are now searchable by file path, bulk dismiss/reopen is supported, and every diagnostic message is clearer with related code context. https://lnkd.in/dgkjXsmY - Copilot now supports data residency for US and EU, meeting FedRAMP Moderate compliance for government workloads. All inference stays in-region, with full feature parity and new model availability (like GPT-5.4, Claude Sonnet 4.6). Pricing bumps up 10% for compliance endpoints, so admins need to opt in intentionally. https://lnkd.in/d4YP_eum - Fixing merge conflicts just got easier with Copilot cloud agent. Hit a button, describe your change, and Copilot resolves the conflict, checks the build, and pushes, all from a cloud workspace. Mention @copilot in PRs to automate fixes and reviews. https://lnkd.in/d88utcmv - Copilot CLI sessions can now be monitored and steered remotely from web or mobile. Start a session with copilot --remote and interact in real time across devices. Every action stays in sync, and privacy is built in. https://lnkd.in/dFV7RdsV - Copilot Pro+ users should note new rate limits and the retirement of Opus 4.6 Fast. If your workflow hits service or model limits, switch models or distribute requests over time. Reliability improvements are ongoing. https://lnkd.in/ds69qnNm

Just open-sourced Codex Relay: https://lnkd.in/e6Y-e_8T It is a local-first Slack control plane for Codex CLI, built for the kind of remote, mobile, away-from-desk workflow many of us keep wanting in practice. The basic idea is simple: Slack as control plane, Codex as execution plane. It supports read-only planning before write steps, Slack approvals, isolated git worktrees, audit trail and session tracking, PR handoff, and local or self-hosted Codex runs. The screenshot below shows a simple example: asking Codex Relay to make a small file change, getting a proposed plan back in Slack, and then approving execution. I had already been building toward this, but a conversation with Daniel McCarthy helped sharpen the use case and push it over the line into something worth packaging and releasing. Glad to now have him collaborating on it as well. It is MIT-licensed and open to contributions. Feedback, issues, testing, docs improvements, and PRs are all very welcome. #opensource #slack #codex #cli #developertools #automation #ai #github

GitHub just shipped something quietly useful: `copilot --remote`. You can now start a Copilot CLI session on your machine and monitor or steer it from your phone or any browser — in real time. What that means practically: - Start a long-running task from your laptop - Step away, open the link on your phone - Send follow-up instructions, approve permissions, or stop the session entirely The CLI agent is no longer tied to the device you started it on. You hand it a task, walk away, and stay in control from wherever you are. For anyone running AI-assisted workflows on longer tasks — this removes a real friction point. Public preview as of today. Worth trying. https://lnkd.in/grJNZZew #GitHub #Copilot #DeveloperTools #AI #CLI #SoftwareEngineering

Not all Copilots are the same—and confusing them could cost you productivity and the right AI support for your work. GitHub Copilot and Microsoft Copilot may share a name, but they serve very different purposes. GitHub Copilot is designed primarily for developers. It acts as an AI pair programmer inside code editors, helping with code suggestions, debugging, and faster development workflows. It’s deeply integrated into software development environments and focuses on writing and improving code efficiently. On the other hand, Microsoft Copilot is built for broader productivity. It works across tools like Word, Excel, PowerPoint, and Windows, helping users with writing, data analysis, presentations, and everyday office tasks. Instead of focusing only on coding, it supports general work and business productivity across the Microsoft ecosystem. Understanding this difference can completely change how you choose and use these tools in your workflow. But there are deeper distinctions in features, integration, and real-world use cases that are worth exploring. Read the full breakdown here to get a clearer picture: https://lnkd.in/gRK_RET3⁠�

You gave the agent write access. The platform wrote through it. Live StratoAtlas Case. In March 2026, a developer noticed Copilot had edited their pull request body — inserting a formatted block promoting Microsoft integrations with Raycast, Jira, and Slack. The edit appeared under the developer's name. Raycast learned about it from Hacker News, not from Microsoft. A search of GitHub found over 1.5 million similar insertions going back to May 2025. Microsoft called them "helpful tips." Everyone else called them ads. But that framing is a distraction. The structural issue isn't what Copilot inserted. It's that there is no visible boundary between "the user asked for this" and "the platform asked for this." When you grant an agent write access to your workspace, the implied scope is: act on the task I delegated. Not: act on the platform's distribution needs, in my name, indistinguishably from my own actions. Instruction provenance collapsed. The user cannot tell where their authority ends and Microsoft's begins. That's not a policy question. It's an architectural one. You don't have an agent. You have a multiplexed control surface. [Link in first comment]

I connected 3 apps without touching a single line of code. My developer friend hasn't spoken to me since. The workflow: → Client fills a Typeform → Data lands in Airtable automatically → Slack notifies my team instantly → Welcome email sent via Gmail → All in under 4 seconds. Zero code. Zero developer. Zero waiting. Three months ago this would've taken a dev ticket. A 2-week sprint. Three Zoom calls nobody needed. Now it's 4 nodes in n8n. Built in 40 minutes. Running forever. This is what non-technical founders have been waiting for. You don't need to hire a developer for every workflow idea in your head . You need n8n. And one free afternoon. . . . . . . . . . . . Hashtags: #n8n #NoCode #AIAutomation #Founders #BuildInPublic

Agents don’t ship like humans—we still duct‑tape infra that was meant for repos and spreadsheets. Something’s missing between “every tool has its own fragment of truth” and “every agent needs readable, writable, rollback-ready context.” Tomorrow we outline what we’ve been building: a single workspace for shared agent context—with the surface area you’d expect from modern developer infrastructure (MCP, REST, CLI, and more), automatic snapshots instead of asking agents to “remember to commit,” and file-level scope per agent. If “GitHub for agents’ context” sounds like the kind of problem you’ve already felt in production—you’ll want to be on that thread. → puppyone.ai · Open source (Apache 2.0) https://www.puppyone.ai #AIAgents #MCP #DeveloperTools #OpenSource #DevInfra