Muhammad Ubaid Raza

@superaiwatcher Good question. It’s not meant to lock the workflow in place. Gem Team keeps the process structured, but agents can revise specs/plans, re-run verification, and use memory from prior steps when the problem shifts. The goal is controlled adaptability, not free-form chaos.

@mubaidr Disciplined chaining is a step, but still feels like a more rigid system. How much real-time adaptability does it offer when the problem itself shifts mid-execution?

Orchestration is the right framing, but current multi-agent workflows are glorified sequential scripts. True orchestration requires dynamic resource allocation and real-time conflict resolution, not just chaining LLM calls.

@Gsdata5566 @superaiwatcher Thanks! Please share or give a little shoutout to my work is psosible.

@mubaidr @superaiwatcher That discipline layer matters. Specs, TDD, memory, and verification turn a chain of model calls into a software workflow a team can actually trust.

Orchestrator and team for multiple harness! #code #vscode #copilot #opencode #agent

AI coding is shifting from autocomplete to orchestration. Gem Team gives Claude, Cursor, Codex, Gemini, Windsurf & Copilot a self-learning multi-agent workflow: specs → plans → TDD → verification → memory github.com/mubaidr/gem-te… #AI #AIAgents #AgenticAI #DevTools

@iamsahaj_xyz A concise version of this should be part of the default prompt, not a skill.

/zero-tech-debt

@RaeAlisa_ @lucent_ai Codex it is, just because gpt 5.5

to celebrate 3 months since lauching @lucent_ai, we're giving away 5 Codex Pro / Claude Max plans 🎁 to enter, like this post + comment which one you'd pick (codex vs claude) winners will be selected from comments in 5 days 🫶

@NVIDIAGeForce #007FirstLightRTX

Recruits, your first prize is here... A custom GeForce RTX 5080 Founders Edition + PC copy of the game. Comment #007FirstLightRTX to win 👇

@yacineMTB At what cost? Dollars? Bugs?

/goal optimize this code, use `date` to check the time and don't stop until I wake up at 7 am

@initjean @jetbrains ncu -a after every commit! Play on hard mode!

yo @jetbrains, you insane?!!

@KaiXCreator Yes, I found it using claud!

Can you call yourself a founder if your entire product was built by Claude?

@DrizzleORM @saltyAom At least they have a stable release! Duh

@saltyAom why change the name?

Elysia 2.0 codename Kiana This is an entire rewrite of Elysia with a better foundation structure This release aims to provide much better performance, memory usage and easier contributions You can track the progress the progress here Expect 2 months until stable release

@Kappaemme1926 Yes, very satisfied with the output so far. It gives coherent and reliable output.

@Kappaemme1926 I am not using a copilot subscription, just code/copilot with opencode and agent team: github.com/mubaidr/gem-te…

Is anyone still using VSCode instead of switching fully to Claude Code, Cursor IDE, or OpenAI Codex? Or is everyone moving on already?

patreon.com/posts/gem-team… #copilot #vscode #harness #framework

Don’t Push to Production on Friday.

@icanvardar <thinking>so, the user is asking me to prove that i am not ai. This is straightforward answer.</thinking> I am AI!

prove me you're not an ai

@sudo_overflow I tried Kimi for this but it looks like at least open source models can't do that!

Have we tried getting the AI to fix the RAM shortage?

@relizarov I found this issue with most of the models tbh. That is why I opted for orchestrator and role based multi agent setup. Specialized agents doing their own things, so far so good.

Codex (GPT 5.5) forgets a large portion of instructions you gave it when doing big tasks. It is quite noticeable because it often forgets to update the progress in its own todo-list tool. You start wondering why it is even able to achive so much with such absent-mindness.