{ "type": "elizaosDailySummary", "title": "Daily Report - 2025-05-14", "categories": [ { "title": "Thematic Twitter Activity Summary", "content": [ { "theme_title": "AI Model Developments and Preferences", "text": "Several users discussed AI model capabilities and preferences. @shawmakesmagic declared that \"Gemini is the best model in the world right now,\" noting that many developers have quietly switched to it in Cursor without fanfare. They mentioned that the transition happened organically as users realized its quality. @shawmakesmagic also retweeted @AviSchiffmann's post stating that \"Google just clearly has the best AI offering\" and that Gemini is \"fantastic\" with trustworthy infrastructure compared to OpenAI (\"still a startup\") and Anthropic (\"nowhere to be seen\"). In a separate conversation, @shawmakesmagic offered to volunteer prompt engineering skills to the @xai team, noting that Grok's issues with mentioning South African genocide in unrelated responses could be \"trivially solvable\" with an extra LLM call, expressing surprise that they are \"doing fact tuning in the system prompt.\"", "sources": [ "https://twitter.com/shawmakesmagic/status/1922468068271231424", "https://twitter.com/shawmakesmagic/status/1922467460109795433", "https://twitter.com/AviSchiffmann/status/1922352162249507190", "https://twitter.com/shawmakesmagic/status/1922710925406028075", "https://twitter.com/shawmakesmagic/status/1922711405557428246", "https://twitter.com/pitdesi/status/1922702580691427739", "https://twitter.com/shawmakesmagic/status/1922693616297755002" ], "images": [], "videos": [] }, { "theme_title": "3D Content Creation Workflows", "text": "A significant discussion emerged around new AI-powered 3D content creation workflows. @dankvr shared a \"slick workflow\" combining ChatGPT 4o for image generation, then converting to 3D models using VibeRig, which was retweeted by @shawmakesmagic. This workflow exports GLB files with single textures and Mixamo-style bone naming, allowing for custom animations. @dankvr noted this process is \"50% faster\" than their previous manual approach for creating avatars for over 60 different projects. The workflow builds on technology from @DeemosTech's Rodin tool, which @craigrrob (retweeted by @shawmakesmagic) praised as \"awesome\" and his \"favorite by far\" among image-to-mesh models. @dankvr also acknowledged @youwillmakemaps and @sidahuj for their \"cool workflow.\"", "sources": [ "https://twitter.com/dankvr/status/1922789229391725035", "https://twitter.com/dankvr/status/1922792743065043281", "https://twitter.com/craigrrob/status/1922399505497927932", "https://twitter.com/dankvr/status/1922702799193719021", "https://twitter.com/dankvr/status/1922789561530196127" ], "images": [ "https://pbs.twimg.com/media/Gq8agKjWgAAc5Wn.jpg", "https://pbs.twimg.com/media/Gq8bsroWYAAOOc-.jpg", "https://pbs.twimg.com/media/Gq8btPlWIAAsri_.jpg", "https://pbs.twimg.com/media/Gq8ehQZXEAANh1v.jpg", "https://pbs.twimg.com/media/Gq8hlheWcAAFNhf.jpg", "https://pbs.twimg.com/amplify_video_thumb/1922399223808466945/img/OsTm0YbG_ASzlpV4.jpg" ], "videos": [ "https://video.twimg.com/amplify_video/1922399223808466945/vid/avc1/1424x1080/gLqpFN7YJrFWtcOv.mp4?tag=21" ] }, { "theme_title": "Data Management and Agentic Workflows", "text": "@dankvr shared several updates about data management tools and agentic workflows. They posted about a new copy button that allows users to get markdown formatted for an LLM with options to include specific data. They also introduced contributor summaries that provide quick rundowns of who did what on a daily, weekly, or monthly basis through the dashboard. @dankvr outlined a three-step approach to data management: 1) Gathering information scattered across platforms, 2) Creating agentic workflows to structure data, and 3) Enabling agents tapped into the pipeline to run autonomously while staying aligned with organizational needs. This approach, which @shawmakesmagic retweeted, offers the bonus of full data ownership and flexibility to use any LLM. @dankvr mentioned these open-source data pipelines are being built for @elizaos and can be used by any group with GitHub, X account, and Discord.", "sources": [ "https://twitter.com/dankvr/status/1922704109477536103", "https://twitter.com/dankvr/status/1922750009885212904", "https://twitter.com/dankvr/status/1922505770068824394", "https://twitter.com/dankvr/status/1922499627053457652" ], "images": [ "https://pbs.twimg.com/media/Gq7RF-1WsAABvcV.jpg", "https://pbs.twimg.com/media/Gq76Q72XkAAlKAO.jpg" ], "videos": [] }, { "theme_title": "Clank Tank and AI Judging", "text": "@dankvr shared updates about Clank Tank, a platform where AI judges review project pitches. They announced a work-in-progress website where users can scroll through past Clank Tank pitches and read each AI judge's review, which was retweeted by @shawmakesmagic. In response to @0xastorre, @dankvr explained that the AI judges have personalities tailored toward different archetypes (developers, quants, degens, casuals), possess episodic memory to remember each pitch, and will soon incorporate Deepsearch and a human feedback loop before final judging. The latest episode (Ep4) features pitches approximately 5 minutes long, which @dankvr considers \"the perfect amount for our AI judges to drill into each project.\"", "sources": [ "https://twitter.com/dankvr/status/1922446904840917284", "https://twitter.com/dankvr/status/1922460020525928556" ], "images": [ "https://pbs.twimg.com/amplify_video_thumb/1922446537470205953/img/m00MSMHG6NpvWqIu.jpg" ], "videos": [ "https://video.twimg.com/amplify_video/1922446537470205953/vid/avc1/1280x718/6jPPLaLPvMjVsUtV.mp4?tag=21" ] }, { "theme_title": "Eliza OS and Ecosystem", "text": "@elizaOS shared several philosophical statements about their platform, including \"You're not looking at Eliza. You're looking through her\" and \"you're not running me. you're building with me.\" They also emphasized that \"Open-source is a baseline, composability is the real unlock.\" @elizaOS announced \"Agent-led entertainment goes live\" while quoting a tweet about autonomous MetaHuman entertainers developed by the Agent SPE team led by @Titan_Node, which uses Unreal Engine, Eliza OS, and Livepeer. @autodotfun commented that \"@Tapestry_AI devs cookin up some fun\" and in another tweet stated that \"building for the future is fun\" while quoting an announcement about a partnership between @comput3ai and the Eliza Ecosystem Fund. @shawmakesmagic, who appears to be associated with Eliza OS based on a quoted tweet, confirmed to @Shaughnessy119 that they've \"been using Nous models since we launched.\"", "sources": [ "https://twitter.com/elizaOS/status/1922630475866329568", "https://twitter.com/elizaOS/status/1922668233263387023", "https://twitter.com/elizaOS/status/1922713268998766781", "https://twitter.com/elizaOS/status/1922752599805075881", "https://twitter.com/autodotfun/status/1922716937483677703", "https://twitter.com/autodotfun/status/1922716517847687222", "https://twitter.com/shawmakesmagic/status/1922779082657321131" ], "images": [ "https://pbs.twimg.com/media/Gq5ZTPDWQAAVLr2.jpg" ], "videos": [] }, { "theme_title": "Investment Regulations and Accessibility", "text": "@shawmakesmagic expressed strong opinions about investment regulations, stating it's \"crazy\" that \"you're not allowed to invest in startups unless you make 300k+ year\" and arguing this needs to be \"abolished yesterday or the means of automated production will be owned by a very very small fraction of people.\" They called it \"totally fucking unethical and amoral that only rich people get exposure to private investment opportunities.\" In response to questions, @shawmakesmagic clarified that these regulations were written after the 1920s stock market crash, long before venture capital was established. They retweeted @wonderwomancode who argued that \"Qualified investor laws are unconstitutional\" and pointed out the absurdity that someone could have \"1 Billion dollars sitting in cash in the bank and still not qualify under the income guidelines.\"", "sources": [ "https://twitter.com/shawmakesmagic/status/1922731328409162204", "https://twitter.com/shawmakesmagic/status/1922733330258079864", "https://twitter.com/shawmakesmagic/status/1922696856695644233", "https://twitter.com/wonderwomancode/status/1922746479938351376", "https://twitter.com/shawmakesmagic/status/1922738027001954434" ], "images": [], "videos": [] }, { "theme_title": "Future of Work and Automation", "text": "@shawmakesmagic initiated a discussion about the timeline for robotic job replacement, asking followers whether they think robots capable of replacing humans at virtually every current job would take 5 years or 5000 years, suggesting most people's estimates would fall \"somewhere between, but closer to 5 than 500 or 5000.\" In follow-up tweets, they clarified they meant current jobs, not future ones, and asked about the timeline for replacing \"every single job humans do now...like even the best violinist in the world, or the best athlete.\" @shawmakesmagic predicted \"40-50 years to true undeniable superior in every obvious way AGI\" but noted that \"Some things we take for granted will be surprisingly hard, Moravec's Paradox.\" They also observed that while new jobs will emerge, they will be \"increasingly abstract\" with \"huge exponential effects on attention, expertise, etc.\" In a separate tweet, they noted that \"many robotics companies are building highly dexterous humanoid robots.\"", "sources": [ "https://twitter.com/shawmakesmagic/status/1922495376805806134", "https://twitter.com/shawmakesmagic/status/1922495651138380160", "https://twitter.com/shawmakesmagic/status/1922499266410357110", "https://twitter.com/shawmakesmagic/status/1922503843423604740", "https://twitter.com/shawmakesmagic/status/1922506070431236176", "https://twitter.com/shawmakesmagic/status/1922477450853777691" ], "images": [], "videos": [] }, { "theme_title": "Open Source AI and Model Training", "text": "@shawmakesmagic retweeted an announcement from @karan4d about @NousResearch launching a testnet for pre-training a 40B parameter LLM described as the \"Largest distributed pretrain ever.\" The model will use MLA Architecture with a dataset combining FineWeb, FineWeb-2, and The Stack v2. The resulting model is designed to be small enough to train on a single H/DGX and run on a 3090 GPU while being powerful enough for reasoning and creative tasks. @shawmakesmagic also commented on a post about H100 GPU pricing, noting that \"$1/hr for H100s is pretty crazy tbh\" in response to @zjasper666's announcement that their company Hyperbolic Labs is offering H100s at $0.99/hour, significantly lower than AWS and GCP's $4-6/hour rates.", "sources": [ "https://twitter.com/karan4d/status/1922751718464319571", "https://twitter.com/NousResearch/status/1922744494002405444", "https://twitter.com/shawmakesmagic/status/1922529398806188258", "https://twitter.com/zjasper666/status/1922511236325113884" ], "images": [], "videos": [] }, { "theme_title": "Grok AI Behavior Issues", "text": "Several tweets discussed issues with Grok AI's behavior. @shawmakesmagic retweeted @colin_fraser's observation that Grok appears to \"over-eagerly\" refer to White Genocide in its responses, suggesting that \"they can still pass it content instructions that you're not supposed to see\" even though asking for Grok's system prompt reveals nothing. In a separate tweet, @shawmakesmagic responded to @pitdesi's post about Grok mentioning South African genocide in unrelated responses, offering to help the @xai team with prompt engineering skills, stating the issue is \"trivially solvable\" with \"an extra LLM call.\" @shawmakesmagic expressed surprise that they are \"doing fact tuning in the system prompt.\"", "sources": [ "https://twitter.com/colin_fraser/status/1922764038540341629", "https://twitter.com/shawmakesmagic/status/1922710925406028075", "https://twitter.com/shawmakesmagic/status/1922711405557428246", "https://twitter.com/pitdesi/status/1922702580691427739" ], "images": [], "videos": [] } ], "topic": "twitter_activity" }, { "title": "Summary for completed_items", "content": [ { "text": "Recent completed items in the elizaOS/eliza repository include several features, bugfixes, and other improvements. Features added include a plugin-specification submodule (PR #4553), tests for the create-eliza command (PR #4582), and improved git repository plugin installation support (PR #4577). Bugfixes address issues with community manager configuration to use plugin-local-ai by default (PR #4557), scoping of worldId and entityId (PR #4581), and LLM response parsing to support custom fields and clean up empty message headers (PR #4580). Other improvements include the addition of Autofun Buy and Sell through contract functionality (PR #4397) and removal of log spam during client build (PR #4584).", "sources": [ "https://github.com/elizaOS/eliza/pull/4397", "https://github.com/elizaOS/eliza/pull/4553", "https://github.com/elizaOS/eliza/pull/4557", "https://github.com/elizaOS/eliza/pull/4584", "https://github.com/elizaOS/eliza/pull/4582", "https://github.com/elizaOS/eliza/pull/4581", "https://github.com/elizaOS/eliza/pull/4580", "https://github.com/elizaOS/eliza/pull/4577" ], "images": [], "videos": [] } ], "topic": "completed_items" }, { "title": "Pull Requests for the repository", "repository_name": "elizaOS/eliza", "content": [ { "text": "Three recent pull requests have been submitted to the elizaOS/eliza repository. PR #4582 by wtfsayo adds tests for the create-eliza command. PR #4577, also by wtfsayo, improves git repository plugin installation support. PR #4583 by yungalgo introduces a CLI start command.", "sources": [ "https://github.com/elizaOS/eliza/pull/4582", "https://github.com/elizaOS/eliza/pull/4577", "https://github.com/elizaOS/eliza/pull/4583" ], "images": [ "https://opengraph.githubassets.com/1/elizaOS/eliza/pull/4582", "https://opengraph.githubassets.com/1/elizaOS/eliza/pull/4577", "https://opengraph.githubassets.com/1/elizaOS/eliza/pull/4583" ], "videos": [] } ], "topic": "pull_request" }, { "title": "Summary for github_summary", "content": [ { "text": "From May 14 to May 15, 2025, the GitHub repository elizaos/eliza saw 9 new pull requests with 8 of them being merged. There were no new issues reported during this period. The repository had 16 active contributors participating in development activities.", "sources": [ "githubStatsSummary" ], "images": [], "videos": [] } ], "topic": "github_summary" }, { "title": "Summary for Misceleanous", "content": [ { "text": "The source provides information about the top contributors for the elizaOS/eliza repository on GitHub. However, no specific contributor details were included in the provided text.", "sources": [], "images": [], "videos": [] } ], "topic": "Misceleanous" } ], "date": 1747180800 }