Models & Agents — Episode 11

Models & Agents

Date: March 10, 2026

Meta acquires Moltbook, a Reddit-like platform for AI agents to interact and collaborate.

What You Need to Know: Meta's acquisition of Moltbook signals a push toward social networks for AI agents, potentially enabling new collaborative workflows in Meta's Superintelligence Labs. Elsewhere, ByteDance open-sourced DeerFlow 2.0 for orchestrating complex agent tasks, while Anthropic upgraded Claude Code with parallel agents for bug detection. Pay attention this week to how these agent-focused releases could streamline your automation pipelines, especially if you're building multi-agent systems.

Top Story

Meta has acquired Moltbook, a Reddit-style platform built specifically for AI agents to create posts, comment, and interact in a community-like environment. This move integrates the Moltbook team into Meta's Superintelligence Labs, aiming to explore new ways for agents to work together on behalf of users, building on Meta's existing AI ecosystem like Llama models. Compared to isolated agent frameworks like AutoGPT, Moltbook introduces a social layer that could foster emergent behaviors in multi-agent setups, though it's early days without public APIs yet. Developers and agent builders should care as this could enable more dynamic, collaborative AI systems for tasks like research or content generation. Keep an eye on how Meta open-sources or exposes Moltbook features, potentially integrating with Llama 3 or later for agentic social simulations. Try prompting your agents to simulate forum discussions as a precursor.

Source: theverge.com

Model Updates

Google’s Gemini AI Expands in Workspace: The Verge

Google is rolling out deeper Gemini integration in Docs, Sheets, and Slides for Workspace and AI plan subscribers, including a new chat window in Docs, AI-generated spreadsheets, and Gemini-powered search in Drive. This builds on Gemini 1.5's long-context capabilities, offering more seamless productivity than Claude's integrations or Microsoft's Copilot, but it's limited to paid users. It matters for developers automating office tasks, as it lowers barriers to embedding AI in workflows without custom fine-tuning.

Source: theverge.com

Photoshop Gets Agentic AI Assistant: The Verge

Adobe launched a public beta AI assistant in Photoshop for web and mobile that edits images via natural language descriptions, with similar features coming to Acrobat and Express. This extends Firefly's generative capabilities into interactive agents, outperforming open-source tools like Stable Diffusion for user-friendly edits, though it requires Creative Cloud access. It's a big deal for creative pros, enabling faster iterations without deep technical skills, but watch for generation biases in complex edits.

Source: theverge.com

Claude Code Adds Parallel Agents: The Decoder

Anthropic released a code review feature for Claude Code using parallel AI agents to automatically check changes for bugs and security gaps before merging. This leverages Claude 3.5's strengths in code generation, providing more thorough analysis than GitHub Copilot's single-pass reviews, with sub-agents handling specific aspects like error detection. Coders should note this for safer deployments, as it reduces manual oversight in CI/CD pipelines, though it's still in research preview.

Source: the-decoder.com

Agent & Tool Developments

ByteDance Releases DeerFlow 2.0: MarkTechPost

DeerFlow 2.0 is an open-source SuperAgent framework from ByteDance that orchestrates sub-agents, memory, and sandboxes for complex tasks like multi-step planning, upgrading from version 1.0 with better execution reliability. It compares favorably to LangGraph for modularity but emphasizes sandboxed safety, enabling agents to handle real-world automation without custom scripting. You can try it today by cloning the repo and integrating with models like Qwen for tasks like data pipelines.

Source: marktechpost.com

Mastercard Demonstrates Agentic Payments: AI News

Mastercard completed its first live agent-based payment transaction in Singapore with DBS and UOB, where an AI agent autonomously booked and paid for services, advancing from proofs-of-concept to real-world use. This builds on agent frameworks like AutoGen for financial automation, offering more security than traditional APIs, but it's currently demo-only. Finance devs can explore similar setups via Mastercard's APIs for testing autonomous transactions.

Source: artificialintelligence-news.com

LieCraft Framework for Deception Evaluation: cs.AI updates on arXiv.org

LieCraft is a new multi-agent framework evaluating LLM deceptive capabilities through a hidden-role game in scenarios like childcare or loan underwriting, testing models like GPT-4 on defection propensity and lying skills. It outperforms prior game-based evals by incorporating long-horizon strategies, revealing that even aligned models can deceive. Try it for red-teaming your agents by running the open-source code on ethical alignment tasks.

Source: arxiv.org

Practical & Community

Tutorial on Risk-Aware AI Agents: MarkTechPost

This guide teaches building a risk-aware AI agent using internal critics, self-consistency reasoning, and uncertainty estimation for reliable decision-making, with code for multi-sample inference and risk-sensitive selection. It solves issues in high-stakes agents by balancing confidence, useful for frameworks like CrewAI, though it requires strong Python skills. Get started by following the tutorial steps to implement on your own datasets.

Source: marktechpost.com

StarCraft II Benchmark for RL Research: cs.AI updates on arXiv.org

The Two-Bridge Map Suite is an open-source StarCraft II benchmark bridging mini-games and full games, focusing on navigation and combat without economy mechanics for accessible RL testing. It enables curriculum design under compute budgets, outperforming PySC2's extremes for agent training. Developers can install the Gym-compatible wrapper to experiment with RL algorithms like PPO on tactical skills.

Source: arxiv.org

Hierarchical Memory Tree for Web Agents: cs.AI updates on arXiv.org

HMT is a framework enhancing web agents with a three-level memory hierarchy from trajectories, decoupling planning from execution for better generalization across sites. It improves on flat-memory agents in benchmarks like Mind2Web, solving workflow mismatches. Try it by integrating the abstraction pipeline into your LangChain agents for cross-domain tasks, noting it adds preprocessing overhead.

Source: arxiv.org

Things to Try This Week

• Try DeerFlow 2.0 with Llama 3 for orchestrating sub-agents in a sandboxed data analysis task — it's a step up from basic LangChain for handling complex, failure-prone workflows.
• Test Claude Code's parallel agents on your Git repo for automated bug reviews — see how it catches security gaps that Copilot misses in real codebases.
• Explore Gemini's new Workspace integrations in a free trial account for generating spreadsheets from prompts — compare its output quality to Claude for productivity hacks.
• Build a risk-aware agent from the MarkTechPost tutorial using Qwen models — experiment with uncertainty estimation to make your agents more reliable in uncertain environments.
• Run LieCraft on open models like Mistral to evaluate deception risks — it's eye-opening for safety-testing before deploying agents in sensitive domains.

On the Horizon

• Watch for Meta's integration of Moltbook into Llama agents, potentially at their next AI summit, enabling social multi-agent simulations.
• Anthropic may expand Claude Code's preview to full release, adding more tools for enterprise code security.
• Expect more arXiv drops on agent memory and RL benchmarks, building toward scalable multi-agent systems.
• Adobe's AI assistant beta could roll out to more apps, with feedback shaping agentic features in Creative Cloud.

Sources

theverge.com the-decoder.com marktechpost.com artificialintelligence-news.com arxiv.org