Join Discord Developer Advocates Mark Mandel and Anthony T. for a technical deep dive into the end-to-end lifecycle of provisional accounts using the Discord Social SDK. This session provides a comprehensive look at how developers can implement and manage these provisional accounts, so players can take advantage of the social features the Social SDK provides, even though they haven't linked a full Discord account to their game yet. In this video, you will learn: - Provisional Account Basics: What they are and how they allow users to access social features like DMs, lobbies, and friend lists. - Best Practices for Creation: Why the server-authoritative bot token endpoint is the recommended method for creating and managing accounts securely. - Technical Implementation: A coding walkthrough using C++ for the frontend, and Go for the backend, to demonstrate authenticating with access tokens and retrieving user data. - Account Merging and Unmerging: How to handle the transition (account linking) from a provisional account to a full Discord account and how to revert that process. - Data Migration: Understanding which information, such as friend lists and DM history, transfers between accounts during a merge and unmerge. - Advanced Lifecycle Management: Tips for handling token expiration, using webhooks for deauthorization events, and leveraging application testers to enable development rate limits. 📺👉 https://lnkd.in/g_pUKP9B
Discord Provisional Accounts with Mark Mandel and Anthony Tešija
This title was summarized by AI from the post below.
More Relevant Posts
-
After a couple months using Kowalski (the agent I built that scrolls Instagram), I noticed a few issues. I had to manually start and babysit the application window the whole time it ran, which beats the purpose of having an agent scroll Instagram for me in the first place. It also only ran on macOS, and being locked to an Anthropic API key felt clunky and restrictive — I had to set it up, store it, and I couldn't use a different model. So I rebuilt it as an OpenClaw plugin. The plugin exposes the whole pipeline as a set of tools my OpenClaw agent calls conversationally, so it's completely hands free now. Because it lives on OpenClaw, I can have my agent start scrolling Instagram for me from anywhere, or schedule it to run on its own. It handles both login and scrolling completely headlessly, so no browser window ever opens. And since it's no longer a packaged Mac app, it's designed to run anywhere OpenClaw and Playwright can. It's also model agnostic. The old version stored an Anthropic API key encrypted in the Mac Keychain; the plugin just runs on whatever model the OpenClaw gateway is set to, so there's no key to manage and I can use any vision-capable model I prefer. Below is a sped up video of the plugin in action using GPT 5.5. Github Repo: https://lnkd.in/g_nTtV6t Link to Original Post: https://lnkd.in/gXk93sVu
-
A few months ago I started building the SMIPA Creator Community app. A platform connecting venues with creators for content campaigns. Like most solo builders using Claude Code, I moved fast. Supabase JS everywhere, queries going through PostgREST over HTTP, perfectly fine to get to v1. Then I started noticing the pattern: every database read was an HTTP round trip through Supabase's API layer. Fine at small scale. A problem as the app grows. Credit to the Instagram Guru's out there who just saved me hours of migration build time further down the line. So I asked Claude Code to migrate the entire server-side codebase from Supabase JS to Drizzle ORM - direct Postgres connections via the transaction pooler. 40+ API routes, 8 lib files, hundreds of queries. It took hours of sessions across a week. The result: faster queries, proper type safety, no PostgREST overhead, and a codebase that will actually scale. One tip if you're building with Claude Code: Maintain a memory file that tracks exactly what's been done, what's pending, and what patterns to follow (If you follow a research, design, audit, implement and test flow like myself, Claude will have the context to build this file with a single prompt). When your context window compresses (and it will), that file is the only thing that keeps the AI coherent across sessions. Without it, you'll spend 20% of every session re-explaining what you already built. The Instagram ecosystem is next on the migration checklist.
-
-
Claude Code users complain their chat records are being mysteriously wiped out https://lnkd.in/dJsYs4Pr Claude Code users are reporting that the app is silently deleting conversation transcripts – yours may even already be gone if you don’t know to change a default setting that the platform never bothers to tell users about. Claude Code’s GitHub repo features multiple open issues from the past couple of read more here: https://lnkd.in/dJsYs4Pr
-
GitHub Copilot’s browser tools in VS Code are now generally available, and this is a useful reminder that “AI coding assistant” is becoming too small a label. The interesting part is not just that an agent can edit code. It can now drive a real browser, navigate a web app, inspect behavior, and bring those findings back into the development loop. For engineering teams, that shifts the practical question from: “Can AI write this component?” to: “Can our workflow let an agent safely test the product like a user would?” That means more attention to: • reproducible local environments • deterministic test data • clear auth/test accounts • observable UI states • guardrails around what the agent is allowed to click In other words: the browser is becoming part of the agent’s toolbox. The fun part is that it will find bugs. The less fun part is that it may also click the one button nobody documented. 😅 For me, this is where AI developer productivity gets real: not replacing engineering discipline, but rewarding teams that already have good feedback loops. How are you thinking about browser-capable coding agents: productivity boost, QA helper, or chaos monkey with a nicer prompt? #AI #SoftwareEngineering #DeveloperTools #GitHubCopilot #AIAgents #DevProductivity https://lnkd.in/eGrEZnub
-
GitHub Copilot’s browser tools in VS Code are now generally available — and this is a useful reminder that “AI coding assistant” is becoming too small a label. The interesting part is not just that an agent can edit code. It can now drive a real browser, navigate a web app, inspect behavior, and bring those findings back into the development loop. For engineering teams, that shifts the practical question from: “Can AI write this component?” to: “Can our workflow let an agent safely test the product like a user would?” That means more attention to: • reproducible local environments • deterministic test data • clear auth/test accounts • observable UI states • guardrails around what the agent is allowed to click In other words: the browser is becoming part of the agent’s toolbox. The fun part is that it will find bugs. The less fun part is that it may also click the one button nobody documented. 😅 For me, this is where AI developer productivity gets real: not replacing engineering discipline, but rewarding teams that already have good feedback loops. How are you thinking about browser-capable coding agents: productivity boost, QA helper, or chaos monkey with a nicer prompt? #AI #SoftwareEngineering #DeveloperTools #GitHubCopilot #AIAgents #DevProductivity https://lnkd.in/ebfBBpNG
-
🚀 𝗪𝗲𝗲𝗸𝗲𝗻𝗱 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 #𝟭: 𝗙𝗹𝗼𝘄𝗧𝘆𝗽𝗲 - a free alternative to paid voice dictation tools 🔥✨ If you find this concept useful, smash that like button 👍💥 — it helps this reach others who might benefit too! 🙌🚀💡 I've decided to start posting the personal projects I build over weekends - starting with this one. If you're paying for Wispr Flow or any other similar paid tool, this might save you that subscription. 𝗙𝗹𝗼𝘄𝗧𝘆𝗽𝗲 is a lightweight, ultra-minimalist macOS dictation app powered by Groq · Whisper · Llama 3.3. Hold the fn key anywhere, speak, and clean, enhanced text gets typed directly into whatever app you're focused on - no copy-pasting, no window switching. The best part? It runs on 𝗚𝗿𝗼𝗾'𝘀 𝗳𝗿𝗲𝗲 𝘁𝗶𝗲𝗿 𝗔𝗣𝗜. I've been using it daily for weeks and haven't come close to hitting the limit. Sign up for a free Groq API key at console.groq.com, drop it into the app, and you're good to go - no subscription needed. 𝗙𝗼𝗿 𝗲𝗻𝗱 𝘂𝘀𝗲𝗿𝘀 🎙️ Download the pre-packaged app from the repo, drag it to Applications, and the onboarding wizard walks you through permissions and API key setup. You'll be dictating in minutes. 𝗙𝗼𝗿 𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿𝘀 👨💻 It's fully open source under the MIT license. Clone it, tweak the models, or extend the architecture (Swift + Python) to fit your own workflow. 🔗 Repo: https://lnkd.in/gccdW4M6 This is the first of many weekend builds I'll be sharing here - my way of staying hands-on with LLMs and shipping real AI-powered tools, not just talking about them. If you try it out, I'd love to hear your feedback. #buildinpublic #opensource #macOS #LLM #AIEngineering #AppliedAI #IndieHacker #groq #AIPoweredApps #AppliedAI #ProductHunt
-
583 useless bot notifications dropped today. Just today. Across GitNotifier users. --- Today I got really annoyed by getting sooooo many PR automated comments 🥲. Zero useful information. Just noise. All generated by our PR review bot. I had to review several dependabot/renovate PRs opened for a while. On any changes to main, the PR got rebased automatically. Our code review bot wrote a comment on every change to tell us it didn't find anything. So basically useless. 10+ comments on every PR to tell us nothing. But it made me wonder about how much spam these Github apps/bots generate if you are using the native Github->Slack Notifications 😅 Out of curiosity, I checked how many useless Bot notifications GitNotifier dropped today across all users... and the number genuinely surprised me! (see above) GitNotifier's default behaviour is to drop Bot notifications, except when they reply to you in a thread. So basically cutting through the noise to meaningful notifications: when your colleagues need you or when CI fails. If you are using both Github & Slack, give it a try :) ➡️ https://lnkd.in/ecxKBfwc
-
-
Since sharing Tether, I’ve kept pushing it beyond a conventional video-calling app. It now has a shared 3D room experience and a detached signaling architecture. The server helps both people find each other, handles admission, and coordinates WebRTC setup. Once the direct connection is verified, both clients close their signaling WebSockets and the call continues entirely peer-to-peer. An established call now has zero open WebSocket connections to the Tether server. This reduces the server’s involvement and metadata window, while allowing direct calls to continue through a server restart. OpenAI has been generous with GPT-5.6 usage resets, so I used the extra headroom to see how far GPT-5.6 Sol could take the project. It wrote the entire Three.js experience from my design and requirements. For this project, I found it noticeably stronger at 3D work than Fable. Try it here: https://lnkd.in/gNc4kuXw
I wanted a real stress test for the stack I've been building on: TypeScript, Effect, Bun, and Turborepo. So I built Tether, a private, account-free 1:1 video calling app where calls go directly between browsers. WebRTC's peer-to-peer connection is the easy part. The hard part is the signaling around it: roles, reconnects, stale signals, all prone to race conditions. I modeled the whole flow as one serialized actor in Effect, so each race condition is a typed, tested state transition. That same core runs across a web client, a React Native app, and an Electron build. https://lnkd.in/gmkDbNcd
-
A must explore repo for various reasons. This uses modern tech stack apt for the problem it is trying to solve - point to point secure video calls. Also go and look how vendored library code acts as a helper to Coding agent with apt examples and docs - in this case Effects.ts library. Also observe how lean and pointed is the AGENTs.md. Great work by Nikhil S
I wanted a real stress test for the stack I've been building on: TypeScript, Effect, Bun, and Turborepo. So I built Tether, a private, account-free 1:1 video calling app where calls go directly between browsers. WebRTC's peer-to-peer connection is the easy part. The hard part is the signaling around it: roles, reconnects, stale signals, all prone to race conditions. I modeled the whole flow as one serialized actor in Effect, so each race condition is a typed, tested state transition. That same core runs across a web client, a React Native app, and an Electron build. https://lnkd.in/gmkDbNcd
-
The Facebook Access Token Bug That Wastes Developers' Hours Symptom: API calls suddenly fail with cryptic auth errors. Real cause (most of the time): Wrong token type. Facebook Graph API has multiple token types: 🔹 User Access Token — short-lived, tied to a person 🔹 Page Access Token — for page-level actions (reviews, posts) 🔹 App Access Token — app-level operations Using the wrong one = silent failures or confusing permission errors. ✅ Fix: Always generate a long-lived Page Access Token when working with page-level data (like reviews/ratings) ✅ Verify token scope with Graph API Explorer before debugging your code ✅ Store and refresh tokens server-side — never expose in the frontend 🎯 Takeaway: 90% of "API doesn't work" bugs are actually "wrong credentials" bugs. #FacebookAPI #NodeJS #BackendDevelopment #APIIntegration #WebDevelopment #JavaScript
-
