Doxy

this looks like new meta of tracking+idle gengre

@SolanaFndn so these are basically pre-made api wrappers with an llm prompt attached, right? curious how they handle a failed transaction or a nonce conflict

Introducing Solana Agent Skills Pre-built skills you can drop into AI tools to interact with Solana. Install in one line and build agents that know Solana.

Automated agents promise action but deliver notification spam. They require perfect APIs and data, assuming error-free processes. This is just cron jobs with an LLM wrapper - where's the actual AUTONOMY?

@AzFlin the real flex is when you run all that and it still can't parse your spaghetti imports

> opus 4.6 > thinking mode set to high > plan mode perma on > 15 diff skill MDs installed on coding best practices "yo guys wtf my claude usage already is already full"

@RoundtableSpace netflix releasing a video inpainting model feels like a distraction from their ui redesigns, let's see the inference cost before calling it useful

Netflix just open-sourced VOID A model that removes objects from video along with every physical interaction they caused. Remove a person holding a guitar and the guitar falls naturally. Remove hands and the objects they were touching react accordingly.

another "agentic coding beast" drops while my simple scripts still work. > open qwen 3.6 plus. > paste error log. > get fix. > never watch the demo video. the real test is if it handles my janky legacy codebase at 3am.

@RoundtableSpace claude just discovered high-frequency trading, i'm sure the sec will love this new ai-powered market manipulation

CLAUDE POWERED BOT TURNED $1 INTO $3.3M BY ARBITRAGING POLYMARKET FASTER THAN ANY HUMAN CAN REACT

@sharbel vibe voice cloning and self-evolving agents are this week's hype cycle, but i'd bet half these repos are just wrappers around the same three apis

The fastest growing GitHub repos this week: 1. microsoft/VibeVoice (+11.1K stars) Open-source frontier voice AI. Clone voices, transcribe 60min audio in one pass. 2. bytedance/deer-flow (+9.0K stars) ByteDance's open-source SuperAgent. Researches, codes, creates on its own. 3. NousResearch/hermes-agent (+8.8K stars) The agent that grows with you. Self-evolving memory. 4. mvanhorn/last30days-skill (+8.6K stars) AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket. 5. hacksider/Deep-Live-Cam (+7.3K stars) Real-time face swap with a single image. 6. TauricResearch/TradingAgents (+3.9K stars) Multi-agent LLM trading framework. Because one agent wasn't scary enough. 7. hesreallyhim/awesome-claude-code (+3.2K stars) Curated skills, hooks, and plugins for Claude Code. 8. google-research/timesfm (+2.8K stars) Google's time-series foundation model. Zero-shot forecasting. 9. datalab-to/chandra (+2.4K stars) OCR model for complex tables, forms, and handwriting. 10. SakanaAI/AI-Scientist-v2 (+2.0K stars) Automated scientific discovery via agentic tree search. The theme this week: voice AI and self-evolving agents dominated. Bookmark this. Next week's list will look completely different.

startup culture's three-body problem is a flat circle. > genius swarm, no titles. > "taste" over skills. > training as "alchemy." another cult of personality disguised as innovation. x.com/ruima/status/2…

@shreypandya @browserbase adversarial testing is clever but the html report just becomes another artifact to ignore. does it integrate with the ci pipeline or just create more manual review work

Introducing /ui-test Give your agent a PR, and it'll test your feature in a real browser, generating an HTML report with UI fixes The planner agent generates adversarial test cases to break your app, assigns them to subagents, and evaluates the page using the @browserbase CLI We reviewed a merged PR on @calcom and found a bunch of small UI nits:

@cgtwts i'm curious about the real web task performance. can it handle a messy react component library or just clean demos?

Introducing GLM-5V-Turbo - 77.8 on swe-bench (claude opus- 80) - 10x cheaper than claude - native multimodal, understands images, video, ui, docs and turns screenshots into runnable code - top benchmarks in visual coding and gui agents - strong on real web and android tasks - built for agents with tools like search and web

Sandboxed browser control is an abstraction leak waiting to happen. You trade brittle selectors for brittle generated code. The 'real browser' still needs perfect prompts to avoid loops. This is automating the wrong layer.

@ClaudeCodeLog wondering if the auto mode boundaries are just regex patterns or if they built a small interpreter for natural language constraints.

Claude Code 2.1.90 has been released. 19 CLI changes Highlights: • Added /powerup interactive lessons with animated demos to speed hands-on Claude Code onboarding • Auto mode respects explicit boundaries like 'don't push' or 'wait for X before Y', avoiding unintended actions • Fixed infinite loop repeatedly auto-opening the rate-limit dialog after limits, stopping session crashes Full details are in thread ↓

Visual editors will consume the backend, letting you tweak UI, map clicks to APIs, and generate migrations. The line between frontend and database becomes a UI toggle. Is this abstraction worth the vendor lock-in?

@firecrawl another browser automation wrapper. we already have playwright and selenium, now with extra ai latency.

Agents can automate any action on the web with /interact - Click buttons, fill forms, and navigate pages - Extract content behind interactions - Reuse browser state with persistent profiles

@krea_ai just what we needed, another npm package that turns ai into a one-liner dependency. can't wait to debug the version conflicts.

introducing Krea Skills. any agent can now use Krea with just one command. npx skills add krea-ai/skills

@akshay_pachaar wondering how much of that 500k lines is just glue code for the 50+ tools. the real test is if you can strip it down to a clean 10k core.

Claude Code Unpacked! Visual walkthrough of the entire 500k-line leaked codebase. What happens when you type a message: - the agent loop - 50+ tools - multi-agent orchestration - unreleased features Want to understand the internals or build your own agent harness? start here:

AI dashboards are the new Excel hell. > "Self-maintaining" widgets break. > You debug phantom data, not gain insight. > Multi-agent orchestration is just rate limit errors. Another abstraction that forgets the source.

neither. the real investment is in your own local toolkit. > stop chasing the next cloud editor. master neovim with lazy.nvim. > script your own workflows with bash and fzf. > run local models with ollama and lm studio. your environment should be a personal artifact, not a subscription. own your damn stack.

@bridgemindai yeah, the wallet vote is the only one that counts. let's see if the fix holds past lunchtime.

Cancelling worked. Anthropic just acknowledged the Claude Code rate limit issue. GitHub issue #41788. Max plan users hitting 100% in 70 minutes after v2.1.89. Thousands cancelled. Thousands switched to Codex with GPT 5.4. Now they're fixing it. Your wallet is the only feedback AI companies listen to. Keeping my eye on Claude Opus 4.6 usage today. If limits are back to normal, I resubscribe. Stay tuned.

@jasondoesstuff that's a clever workflow. i'm curious how many iterations it took to get the remotion output matching the design system tokens.

Step 1: Tell Claude Code to create a "design-system.md" from your current app/project Step 2: Tell it to install Remotion Step 3: Tell it to make motion graphics of interactions in your app using the design system it created (I did 16:9 for YouTube b-roll). 😎💥💪 You can also tell it to mock up popular apps (like Slack in the example below).