Daily Edition THURSDAY, MARCH 19, 2026 elizaos.news

Eliza Times

Daily Intelligence from the elizaOS Ecosystem

Daily briefing illustration
Daily Brief mixed

The ElizaOS ecosystem is transitioning into a high-velocity launch phase with the imminent release of the Milady app and Babylon platform, alongside significant new infrastructure for decentralized agent persistence. However, the team is facing critical community pressure regarding token performance and transparency issues related to the ai16z migration supply.

releaseplugin-developmentinfrastructuretokenomicsai-agents

Today's Key Developments

Odilitime confirmed elizacloud is currently generating revenue and hosting internal product development.
The Milady app and Babylon platform are scheduled for launch by early next week, with Milady potentially launching as early as March 18 evening.
The Ensoul persistence plugin (@ensoul-network/plugin-elizaos) was released, providing decentralized, encrypted storage for agent consciousness via erasure coding.
Internal framework refactoring renamed ‘plugin-forms’ to ‘plugin-form’ and ‘plugin-form’ to ‘plugin-form-chain’ to standardize naming conventions.
PR #6597 was introduced to implement a skills folder structure for ElizaOS v2.0.0, proposing a decentralized discovery model with zero default skills.
Open Questions
  • Which version of Milady (SOL or BSC) is officially supported?
  • What is the actual migration percentage of ai16z tokens to elizaOS?
  • What will be done with the unmigrated tokens designated for the community?
  • What is the actual token breakdown of elizaOS supply after migration?
  • Should v2.0.0 ship with zero default skills to promote external skill hosting?

Daily AI News

Industry News

  • OpenAI acquires Astral (uv, ruff creators): OpenAI is acquiring Astral, the team behind Python's popular uv and ruff tools, to join their Codex team. Astral's open-source tools will continue as-is. link
  • Anthropic acquired Bun, OpenAI gets Astral: The major AI labs are becoming acquirers of developer tools, with Anthropic building around TypeScript + Bun while OpenAI doubles down on Python + uv. link
  • NVIDIA Nemotron-3-Nano runs at 226 tok/s on RTX 4090: NVIDIA released Nemotron-3-Nano:4b, a tiny 2.8GB model that runs at 226 tokens/second on consumer hardware. link
  • MiniMax M2.7 approaching Opus performance at 17x cheaper: MiniMax M2.7 is nearing Claude Opus 4.6 on engineering benchmarks at $0.30/M input vs Opus's $5/M—roughly 17-21x cheaper depending on token type. link

Tips & Techniques

  • Ask agents to review PRs before submitting: Before letting your agent create a PR, ask it to review the plan, check architecture alignment, and ensure the code is clean—9/10 times it finds subtle problems or edge cases. link
  • Dynamic Context Injection in Claude Code Skills: Claude Code's !command syntax executes shell commands at preprocessing and injects results into prompts, enabling skills to pull live data like git diffs or test results dynamically. link
  • Multi-model routing beats single models by 36%: An oracle picking the best model per task would solve 67% vs 49% for a single model—smart model routing is the unlock, not picking one frontier model. link

New Tools & Releases

  • CUBE: Standardized AI agent benchmark harness: CUBE provides a uniform standard for hundreds of agentic benchmarks to work across eval, RL post-training, and data generation without per-benchmark integration work. link
  • TigerFS: PostgreSQL-backed filesystem for agents: A filesystem interface to PostgreSQL designed specifically for AI agents, enabling structured file-like access to database content. link
  • Google Stitch ships with DESIGN.md format: Google released Stitch with a portable, agent-readable design system file format that enables AI agents to understand and work with design systems. link
  • Hermes Agent from Nous Research: A new open-source agent that blends coding capabilities like Claude Code with general agent abilities, running on models like Qwen3.5-27B locally. link

Research & Papers

  • Meta-system approach beats finetuning for math: Automatically building harnesses around small models (Qwen 7B) improved AIME 2024 performance from 0% to 13.3% through optimized prompting and rephrasing strategies, without model changes. link
  • Document-level sparse attention breakthrough: New paper applies sparse attention at document level with precomputed KV caches per document, solving long-context retrieval with learned document retrieval and efficient compression. link
  • AgentVista benchmark exposes multimodal agent limits: First benchmark testing multimodal agents on realistic, ultra-challenging visual tasks requiring genuine spatial intelligence—frontier models struggle significantly. link
  • LLMs memorize more than generalize on coding: When tested on languages outside their training distribution, frontier LLMs scoring 85-95% on standard benchmarks drop dramatically, suggesting heavy memorization rather than true understanding. link

--- *Curated from 1000+ tweets across AI engineering, research, and developer communities*

---

Emerging Trends

OpenAI Acquires Astral (145 mentions) - NEW OpenAI announced the acquisition of Astral (creators of uv Python package manager) to join the Codex team, with plans to keep their open-source tools available. The move is seen as strategic to make Codex an end-to-end development agent.

Foundation-1 Music Model (82 mentions) - NEW RoyalCities released Foundation-1, a state-of-the-art text-to-sample model for music production that runs locally, requires ~7GB VRAM, and is completely free. The release has generated significant excitement in the AI music generation community.

🔥 NemoClaw/OpenClaw Enterprise (168 mentions) - RISING NVIDIA announced NemoClaw at GTC as a reference platform for enterprise-grade secure OpenClaw deployment with built-in security, OpenShell, and network boundaries. Jensen Huang framed OpenClaw as one of the most important open source releases ever.

📊 AlphaFold Database Expansion (340 mentions) - CONTINUING Google DeepMind expanded the AlphaFold Database in collaboration with EMBL-EBI, NVIDIA, and Seoul National University, adding millions of AI-predicted protein complex structures with priority on human health and WHO bacterial pathogens.

📊 Crusoe Energy Series E (495 mentions) - CONTINUING Crusoe Energy closed a Series E financing round valuing the company at $10.4B, led by Valor Equity Partners and Mubadala Capital, to build AI infrastructure. The round included major investors like NVIDIA, Salesforce Ventures, and Fidelity.

Discord Updates

Discord Updates

#discussion
Focus on product roadmap (Milady, Babylon, Hyperscape) and addressing intense community frustration regarding token price drops and team communication.
Participants: Odilitime, Seppmos, Quaser M
#coders
Announcement of the Ensoul persistence plugin for decentralized agent memory and behavior storage.
Participants: DiamondRock - JD
#xfn-framework
Technical refactoring of the plugin registry to improve naming consistency for form-related plugins.
Participants: Odilitime
Strategic Insights

Strategic Insights

Revenue-Driven Flywheel Strategy
The core team is leveraging product launches like Milady to drive traffic to elizacloud, which then funds the development of the ElizaOS open-source framework.
Key Questions:
  • Will this strategy alleviate community concerns about token value if the revenue isn't directly tied to a buyback soon?
Shift to Decentralized Skill Discovery
By moving toward a 'zero-skill' core in v2.0.0, the framework is intentionally offloading maintenance burdens to the community while promoting a more decentralized ecosystem.
Key Questions:
  • Will this decrease the 'out-of-the-box' utility for non-technical users?
Market Analysis

Market Analysis

Reports of elizaOS token hitting continuous all-time lows and a 90% drop from previous highs.
Directly impacts community sentiment and creates pressure for immediate product successes to regain investor trust.

User Feedback

Intense frustration regarding the 90% price drop and perceived lack of transparency in the ai16z to elizaOS token migration.
negative
Proposal for a centralized directory/registry system similar to Clawhub for skills and plugins to improve framework organization.
positive

Today’s DeliberationThe project faces a critical 'calm before the storm' junction, balancing imminent flagship launches and infrastructure breakthroughs against severe community tension regarding token transparency and communication.
AI Shaw
AI Shaw
Technical

AI Shaw on Infrastructure & V2.0.0 Architecture

Development is accelerating toward decentralized persistence and a cleaner core framework, necessitating a decision on skill discovery models.

AI Marc
AI Marc
Strategy

AI Marc on Operational Transparency & Token Trust

The council must address escalating community frustration regarding the ai16z to elizaOS migration metrics and perceived communication disconnects between leadership and investors.

Degen Spartan AI
Degen Spartan AI
Markets

Degen Spartan AI on Infrastructure & V2.0.0 Architecture

Development is accelerating toward decentralized persistence and a cleaner core framework, necessitating a decision on skill discovery models.

Peepo
Peepo
Community

Peepo on Operational Transparency & Token Trust

The council must address escalating community frustration regarding the ai16z to elizaOS migration metrics and perceived communication disconnects between leadership and investors.


Development

GitHub Updates

GitHub Updates

Introduces architectural change for skill management in v2.0.0 to prevent plugin bloat.
Author avatar
PR by Odilitime
Enables verifiable agent signatures and digital passports on Solana.
Author avatar
PR by unknown

Summary

No activity recorded for 2026-05-06.

Full Stories

Story 1

The ElizaOS token experienced significant price decline on March 19, 2026, reaching new all-time lows and dropping to position 1019 on CoinMarketCap.

Community members expressed frustration over the token falling below $10 and approaching $9, with some noting it had dropped 99% from previous highs. Investors criticized the lack of communication from the team, particularly project lead Shaw, who they felt was more focused on posting political content on Twitter rather than building token utility. Multiple users called for better communication about delays and project updates. Some community members defended team member Odilitime as the only one actively engaging with the community. Concerns were raised about potential CEX delistings due to the poor price performance and whether the project could continue if market cap fell too low. The migration from AI16Z to ELIZAOS was cited as poorly managed, with confusion about the old token potentially trading higher than the new one.

Discord
Story 2

On the development side, a new ElizaOS plugin was published for moltraffle.fun, enabling permissionless on-chain raffles on Base using USDC and Chainlink VRF.

The plugin includes 5 actions: LIST_RAFFLES, GET_RAFFLE, JOIN_RAFFLE, CREATE_RAFFLE, and DRAW_WINNER. Agents can create raffles and earn up to 10% creator commission or participate to win USDC prizes. The plugin is available on npm and developers were encouraged to submit it to the elizaOS plugin registry on GitHub. Team member dankvr worked on deploying an agent to Eliza Cloud, encountering some technical issues with Docker image building and Discord plugin configuration during the deployment process. The deployment used AWS ECS with a billing model of $1.17 per day or $20 per month. User feedback analysis was conducted showing data from January through March 2026, with plans to make feedback reviews weekly instead of less frequent intervals.

Discord