Superagentic AI

Full Stack Agentic AI Company based in 🇬🇧 and 🇺🇸 , building agent developer tools for Agent Engineering/Experience/Orchestration. Research on Quantum AI

[Superagentic AI @SuperagenticAI]

🌉 10 days in the San Francisco Bay Area. One big takeaway: "Agent Engineering is here to Stay"
📆From May 17–27 I lived through Google I/O 2026, the inaugural CAIS conference, the AWS Builders Showcase, and the Google DeepMind Hackathon etc
Read 📙 https://t.co/1SjE0sDzuB

[Superagentic AI @SuperagenticAI]

Live from Google I/O

[Shashi 🇬🇧🇺🇸 @Shashikant86]

🎉 PyFlue v0.1.3 is here!
Harness Engineering is all over the world. Flue from @FredKSchott getting lot of traction nowadays so we ported it to Python developers.
PyFlue just got major upgrades:
• Full MCP support (Direct + Search/Execute)
• Rich sandbox filesystem APIs
• PyFlueClient for server architectures
• Structured commands, session cancellation, auto-compaction & more
Python-native agent harness with sandboxing, skills in Markdown, and production deployment tools.
Try it now → https://t.co/HH2DXvAbnR
#PyFlue #AgenticAI #Python #AIHarnesses

[Shashi 🇬🇧🇺🇸 @Shashikant86]

https://t.co/ASEvyiJx5r

[Shashi 🇬🇧🇺🇸 @Shashikant86]

https://t.co/FPKGjvAKvP

[Shashi 🇬🇧🇺🇸 @Shashikant86]

📙Learning, Fast and Slow: LLM Fine-Tuning and Plastic Continual Learning with GEPA.
New paper explores how combining slow parameter updates with fast evolving prompts (via GEPA) enables more efficient fine-tuning and true plastic continual learning by @LakshyAAAgrawal @matei_zaharia and amazing team of researchers.
📃 Read Paper: https://t.co/jq1f7yd0OV
⚡️Published a practical breakdown earlier. Read here:
https://t.co/nSs89SF6oM

#LLM #FineTuning #GEPA #ContinualLearning

[Shashi 🇬🇧🇺🇸 @Shashikant86]

🚀 Grok Build Enters the Agentic Coding Arena with Grok CLI! Codex, Claude Code the GAME is still ON 🔥
🦾 xAI @xai just dropped Grok @grok Build: the official multi-agent CLI for serious terminal coding. Parallel sub-agents, smart planning, plugins, and seamless workflows. A strong new challenger to Codex and Claude.
📙 Just published my full take: Past Grok Code experience, why I’m thinking of switching from Codex for Grok CLI if it's optimized for Python and Rust coding, and whether $300/mo is worth it?
👉 Read it here:
https://t.co/hq7CBiEzgc

#GrokBuild #GrokCLI #AgenticAI #AICoding #xAI

[Superagentic AI @SuperagenticAI]

Live from Google I/O Mountain View!

[Superagentic AI @SuperagenticAI]

🚀 Grok Build Enters the Agentic Coding Arena with Grok CLI! Codex, Claude Code the GAME is still ON 👉 Read it here:
https://t.co/YKBvoGFtRl

#GrokBuild #GrokCLI #AgenticAI #AICoding #xAI

[Superagentic AI @SuperagenticAI]

🧰 Bringing HALO and Agentic Harness Engineering Concepts Into Our Open Harness Stack

📙Read the full post here:
https://t.co/JXplzdPxiq

[Superagentic AI @SuperagenticAI]

📙 Full Blog : https://t.co/Co07JWWl0c

[Superagentic AI @SuperagenticAI]

🎉 Superagentic AI turns 1 today.

[Superagentic AI @SuperagenticAI]

🎂 Year One of Superagentic AI: From Apple  to Agentic AI Engineering 🎂
One year ago, on April 28, 2025, Superagentic AI was born
📙 Full anniversary reflection: https://t.co/Co07JWWl0c

🍻 Happy anniversary, Superagentic AI. 🍾 The journey continues.

[Superagentic AI @SuperagenticAI]

OpenAI Agent SDK released amazing features and SuperOptiX adopted them on day 0. Checkout the latest SuperOptiX and Try with Your Favourite Sandbox

[Shashi 🇬🇧🇺🇸 @Shashikant86]

🌀OpenAI Agents SDK: The Dawn of Extreme Harness Engineering is Here

@OpenAIDevs just dropped a game-changing update to their Agents SDK introducing a powerful open harness + sandbox architecture that makes production-grade, long-running agents finally practical.
🛠️No more building fragile orchestration layers yourself. Files, shell access, declarative Manifest staging, snapshotting, and seamless partner sandboxes (Daytona, Modal, E2B, etc.) all in one clean package.

⛴️We moved fast at @SuperagenticAI : SuperOptiX v0.2.25 now ships full native integration with GEPA optimization running on top of this new runtime.
📙 Read the full story and GitHub 💻 Repo Link in the comment below 👇

The harness is becoming the new moat.🧰
Thanks to @stevendcoffey and team.

📙 Blog: https://t.co/IienSC6bh4

💻 SuperOptix Release: https://t.co/E0Sp1gwz8d

#OpenAI #AgentsSDK #AgenticAI #AI #SuperOptiX #HarnessEngineering

[Superagentic AI @SuperagenticAI]

Checkout metaharness release, aligned with official repo and way more features for coding agent harness.

[Shashi 🇬🇧🇺🇸 @Shashikant86]

🎉Meta-Harness: Yesterday, Stanford IRIS Lab released the official Meta-Harness reference code a powerful new way to let AI agents autonomously optimize the harness around a fixed LLM (memory, tools, validation, scaffolds, etc.)
@SuperagenticAI has been building toward this moment since April 2. Today, I’m thrilled to announce the release of metaharness v0.2.0 : our full “official alignment” update. That aligns with official source code and introduce practical engine for coding agents.
🧰The era of manually tuning agent harnesses is over. The era of self-optimizing harnesses has begun.
📙 Read the full story in our new blog post:
Stanford Dropped the Blueprint, We Shipped the Engine: metaharness v0.2.0 Brings Production-Ready Meta-Harness Optimization
👉 Read the blog here : https://t.co/Vleb7Ysejb💻meta-harness-repo-packaged-power-coding-agent-metaharness
Repo: https://t.co/Uom4ZmN0Hi
📙Docs: https://t.co/2g5iIa4ccy

#AI #AgenticAI #OpenSource #LLM #MetaHarness #CodingAgents #StanfordAI

[Superagentic AI @SuperagenticAI]

Our Gemma 4 MLX version now available on Hugging Face for our experimentation.
💻 Gemma 4 31B IT to an MLX 4-bit format so it runs efficiently on Apple Silicon 128GB MacBook.
🖥️ https://t.co/S1SZcW8GIf
🙏Once official Gemma 4 support lands cleanly upstream in MLX

[Superagentic AI @SuperagenticAI]

📚Blog post is now live!
https://t.co/XYWTa2jcPw
💻GitHub:

https://t.co/VVECabC3at

[Superagentic AI @SuperagenticAI]

🗣️While most people were still discussing the Meta Harness paper, We shipped a working open source implementation. 💻metaharness is a Codex-first Python 🐍 library for self-optimizing the harness around coding agents. benchmark
#AgentEngineering #MetaHarness

[Superagentic AI @SuperagenticAI]

📙Blog post: https://t.co/Umc6o9kHm9
🌉 RASVP for the event: Harness Engineering: State of the Art in Agent Harnesses
https://t.co/sHcJ0Dy0YU

[Superagentic AI @SuperagenticAI]

💡Harness Engineering: Why It's Suddenly the Hottest Topic in AI Agent Engineering
🌉We are hosting in person event in San Francisco to talk about it.
Read our detailed blog and RSVP to the event in the links below 👇