lee.aki

Blockchain Builder | Holder of #BTC #ETH

[Jerry Liu @jerryjliu0]

We built the fastest PDF -> markdown parser in the world 🚀⚡️

AND it’s more accurate than any other open-source, model-free parser (pymupdf4llm, opendataloader, pdf-inspector, markitdown) on 3 standardized benchmarks: olmOCR0-bench, opendataloader-bench, ParseBench

Introducing LiteParse v2.1. The v2 base version was already the fastest document->text parser on the planet, and with this new release we’ve introduced markdown.

It is fully open-source (Apache 2.0) and free, is usable from CLI/Rust/Node/Python/WASM, and is also installable as a one-click agent skill.

Check it out: https://t.co/7oFImAZeb2

Come check out LiteParse: https://t.co/JNER0mVcB8

[Addy Osmani @addyosmani]

https://t.co/hIe0UX7z6T

[Jerry Liu @jerryjliu0]

We've created the world's fastest PDF parser ⚡️

And it's more accurate than any other open-source, model-free PDF parser out there (pymupdf, pypdf, markitdown, pdftotext, opendataloader, pymupdf4llm)

Introducing LiteParse v2 - we rewrote the entire library into Rust and adapted it as native packages for Python and Node.

It supports 50+ different document types, can be triggered directly or installable directly within your favorite AI agent.

Blog: https://t.co/ckb0G73ESs

Repo: https://t.co/JNER0mVcB8

[lee.aki @leeaki_eth]

@HuyLe_HC @DecibelTrade Ai mua giá này không để bán :))

[George from 🕹prodmgmt.world @nurijanian]

if you want to design with AI agents, these skills are amazing

- impeccable https://t.co/Wcykv4uHwT
- taste https://t.co/rThOMA1z76
- layers https://t.co/VKFrJiCoBN
- superdesign https://t.co/cibZkPsd3D

I also made a plugin based on Refactoring UI (use for polish): https://t.co/hV6iVpBgqf

find more here: https://t.co/5LJs9brXDv

[KAWAI @kawai_design]

【 文字のアニメーション100選 】

Webサイトに動きが欲しい。
でも、何て指示すれば良いの？

そんな疑問を解決するサイトを作りました。アニメーション名・デモ・コードコピーを100個掲載しています。もうアニメーションを指示する際に迷うことはなくなると思います。（100個から1個を選ぶ大変さはあるかもw）

🔽 Webサイトはリプに掲載

[Propr @ProprXYZ]

We have three new open roles.

- Software Engineer
- Quant (Risk)
- Affiliate Manager

$10,000 referral bonus for the engineering hires. Apply below.

[Minara AI @minara]

Quant trading is solved.

Introducing: Minara Strategy Studio.

Design strategies, backtest, compare PnL, avoid brutal drawdowns, optimize parameters, go live, all in plain words.

Join the closed beta: https://t.co/RClebUHKm1

🧵

[Propr @ProprXYZ]

Prop firms profit from slippage they never show you.

Every fill on Propr displays a breakdown with best bid, fill price, and slippage calculated by walking Hyperliquid's order book at execution.

If your prop firm hides slippage stats, ask why.

[Thariq @trq212]

https://t.co/Ljzw0lmvao

[self.dll @seelffff]

i cancelled $2,000/month in trading subscriptions

replaced every single one with open-source repos
here's the full stack:

1. TradingView Pro ($30/mo) → lightweight-charts
   14K stars. by TradingView themselves. 45KB. free
https://t.co/Zj8BoF0kbj

2. Bloomberg Terminal ($2,000/mo) → fredapi + Claude
   every macro dataset the Fed publishes. free API
https://t.co/QOsmACH9tB

3. backtest platform ($100/mo) → prediction-market-backtesting
   NautilusTrader fork with Polymarket + Kalshi adapters
https://t.co/ezKB2PSBUq

4. real-time dashboard → polyrec
   terminal UI: Chainlink oracle, Binance feed, orderbook depth
   70+ indicators. auto CSV logging. strategy backtester
https://t.co/fYj5aFUTS4

5. bot framework (7 strategies) → Polymarket-Trading-Bot
   53K lines TypeScript. arbitrage, momentum, market making,
   AI forecast, whale copy-trade, convergence
https://t.co/xNSLjlIEZd

6. strategy reverse engineering → polybot
   execution + market data infrastructure. paper trading
   Kafka, ClickHouse, Grafana. full analytics pipeline
https://t.co/s3fjSwXV6z

7. paper trading for AI agents → polymarket-paper-trader
   real order books. exact fee model. slippage tracking
   your Claude agent gets $10K paper money and trades
https://t.co/oXMxD9uhKI

8. token savings → rtk
   CLI proxy. cuts Claude Code tokens by 60-90%
   Rust. single binary. 10 AI tools supported
https://t.co/WKnP7dfvuj

9. Claude Code itself ($200/mo) → goose
   35K stars. by Block (Jack Dorsey). Rust
   works with any LLM. full agent loop. free
https://t.co/md2P9CJ4Ia

10. wallet tracking + copy trading → Kreo
    track top Polymarket wallets. auto copy trades
    the only tool on this list i actually pay for
    because it makes more than it costs
https://t.co/rVKQ107tBV

total before: ~$2,600/month
total now: $0 + Kreo

bookmark this. you'll need it

[Peter Steinberger 🦞 @steipete]

The next version of @OpenClaw comes with native video generation. To start, I added support for the following companies:

- Alibaba
- BytePlus
- fal
- Google
- MiniMax
- OpenAI
- Qwen
- Together
- xAI https://t.co/NeSp4shVEx

[Peter Steinberger 🦞 @steipete]

@BlackOpsREPL I agree, but doing what I can to soften the fallout.

You can also use the cli as model provider:

models auth login --provider anthropic --method cli --set-default

[jack @jack]

https://t.co/Ev47RckJH5

[Alex @alexgilev]

Agentic UI is LIVE 🚀
https://t.co/v4JPVorbYj

✅ 40+ unique component sets
✅ 700+ total variants
✅ Light + Dark mode
✅ LLM-ready
✅ Style tester for UI
✅ Variables (Color, Type, Spacing)

[Nick Babich @101babich]

🔥 Figma Skills for Claude Code

Figma recently released a collection of Skills for Claude Code that allow Claude to take actions on the Figma canvas.

🔺 What exactly are Figma skills

Think of Figma skills as a collection of instructions that teach Claude Code how to interact with Figma and perform specific actions. Skills are [dot]md files (md stands for markdown) that contain the instructions.

🔺 How to install Figma Skills in Claude Code

The right way to install Figma skills is to install the Figma plugin, because the skills are included automatically with that setup. You can do it with the following command:

claude mcp add figma

🔺 What Figma skills can we use

There are two types of Figma skills: foundational and task-specific.

[slash]figma-use is an example of a foundational Figma skill. This skill teaches your Claude what Figma tokens, variables, styles, and components are and how to use them on the canvas. Most of the time, this skill is triggered automatically when Claude Code tries to perform tasks in Figma.

Task-specific skills are skills that are triggered on demand, either by you manually or by Claude Code automatically.

✅ 5 Examples of task specific Figma skills you can use

1️⃣ [slash]figma-generate-design : Create new designs in Figma using existing components and variables.
2️⃣ [slash]igma-generate-library : Create new components in Figma from a codebase.
3️⃣ [slash]apply-design-system : Connects existing Figma designs to published design system components.
4️⃣ [slash]audit-design-system : Audits Figma screens for design-system integration drift.
5️⃣ [slash]rad-spacing : Applies hierarchical spacing with variables and fallbacks.

📕 Complete guide to Figma skills below 👇

[Haseeb ＞|＜ @hosseeb]

Claude Code's source code just got leaked, so I used Claude Code to study its own source code and compare it against Codex.

Interesting finds:

1. Claude Code has multiple silent mini-compaction events before global compaction kicks in. It silently compacts previous turns and tool calls to manage context size, to extend runway before you get stop-the-world compaction. This is a lot of the secret sauce of why long sessions feel so good.
2. Anthropic has special custom prompting changes that only activate for Anthropic employees. This includes: "Never claim 'all tests pass' when output shows failures." and "Keep text between tool calls to <=25 words." Worth adding these to your global claude.md, as they almost certainly have a ton of data that these improve performance.
3. Claude Code precomputes responses while the user is typing, to make response times feel magical and snappier. For simple responses like "yes," the harness starts processing your likely result before you ever hit enter.
4. Claude Code's interface is actually written in React, but then rendered to the terminal. Not typical for TUIs. Big part of why the memory footprint is so high?

[V @vineetwts]

I created documentation over Claude Code's Codebase, which explains

- Its pipeline
- How it works
- How it handles Context
- How it handles Memory
&amp; More

Read it here - https://t.co/GngrSvWAmh

[Viktor Oddy @viktoroddy]

Google Stitch vs Claude

Prompt 👇

[Uttam @uttam_singhk]

🚨 Tempo & Stripe just launched MPP which is a new open standard internet-native way for agents to pay and if you've been following x402 standard, this is gonna be v interesting

MPP is more of a superset of x402 imo

x402's exact flow maps directly onto MPP's charge intent, so existing integrations carry over but there are additions to it

What MPP adds:
→ Payment-method agnostic: stripe cards, @tempo stablecoins, custom rails, not just stablecoins on blockchain

→ Flexible billing models: one time charges, pay-as-you-go sessions, per token streaming, and payment channels, not just per-request

→ Agent native: purpose-built for LLM agents — session intents pre-authorize budgets for
sub-100ms latency and near-zero per-request overhead

→ Production-grade primitives: cryptographic challenges, credentials, and receipts exchanged over standard HTTP headers or MCP/JSON-RPC bindings

FYI there is no Facilitator component