An Insider's Account: I Worked at Three Exchanges That All Imploded? A former employee of WebSea, JuCoin and CoinUP told WuBlockchain that some small and mid-sized exchanges rely on rising platform-token prices, discounted token purchases, high-yield products, copy trading and agent networks to attract user funds and keep them within the platform. He said WebSea allowed employees to withdraw while ordinary users were unable to do so during a 2024 crisis; JuCoin targeted users in lower-tier markets through mining products and a complex multi-level agent system; and some CoinUP team leaders or agents were required to purchase its CPX platform token, which later suffered a severe price collapse. He said ordinary users could face withdrawal restrictions while employees, agents and other insiders could sometimes withdraw more easily. At JuCoin, withdrawals exceeding deposits were less likely to be approved and profits could be deducted. Dougao argued that in a bear market, some smaller exchanges may increasingly turn to platform tokens, high-yield products and lower-tier user acquisition, highlighting the risks behind “principal protection,” high returns and discounted token purchases.

目前我的完整~/.claude/settings.json { "permissions": { "defaultMode": "bypassPermissions", "allow": [ "Bash(*)", "Read(*)", "Edit(*)", "Write(*)", "WebFetch(*)", "WebSearch(*)", "Glob(*)", "Grep(*)", "NotebookEdit(*)", "Agent(*)" ], "deny": [ "Bash(rm -rf /)", "Bash(rm -rf /*)", "Bash(git push --force origin main)", "Bash(git push --force origin master)" ] }, "attribution": { "commit": "", "pr": "" }, "hooks": { "PreToolUse": [ { "matcher": "Write|Edit", "hooks": [ { "type": "command", "command": "~/.claude/hooks/check-secrets.sh" } ] } ] }

While AI Safety debate is ranging, let me change the topic slightly and share something phenomenal I discovered today. 1) Importance of WebSearch If you replace DeepSeek Harness' default search with @ExaAILabs , your performance of DS V4.1 Flash model will be consistently at Astra High/Fable level. @ExaAILabs rocks! Before this, I did not realize Search API can make so much difference to quality of answers and analysis. I thought it to be an undifferentiated product. Earlier during my testing at least 10% of the model used to answer dumb, while other 90% of times it used to be good. But, now it is consistently great. At its max setting, the model is at Astra High level - which is very astonishing. @teortaxesTex @WilliamBryk 2) Desire to blow your mind: I was asking a complex tech question to DS V4.1 Flash and I was disagreeing with the model (it was about Huawei's 7.2T NPO Engine specification). It held its ground and it provided this diagram to me. It was pretty good for the answer. I though how amazing Exa is (which it absolutely is), then I realised this diagram was made by DS V4.1 Flash by cutting and pasting diagrams from different sources and adding the red coloured markers. It was f*cking precise!!! This model has so much potential for greatness, on top of being exceedingly fast. These experiences have changed my viewpoint towards flash models. May be flash is all you need! @zephyr_z9 @jukan05 @vikramskr @iamfabian @austinsemis 3) Fast Fast Deep Deep Research: I really stopped ChatGPT and Claude's Deep Research because - although it was excellent - it used to take lot of time. With DS V4.1 Flash running nearly at 300+ tokens per second and fast Exa search, I can use DeepSeek Harness's PTC mode (programmatic tool calling) or Agent Team mode (launches subagents) and get very well done research that is very fast (although later is a bit flaky). Astra Max is something that is industry leading in my opinion, but DS V4.1 Flash with DS Harness is exceptional (cost and speed wise wrt to quality ration is best possible)

Ghast AI v0.1.9 is live. 🆙 ♻️ Improved Loop stability and recovery for long tasks. 🛠️ Added user-approved crash diagnostic uploads. 🔍 Strengthened WebSearch safety with stable fallback paths. 🧩 Raised the Skill learning threshold to reduce low-value Skills. 🤖 Added personal AGENTS instructions for chat and coding tasks. try it👇

# OpenCode Features and Practical Usage 🧰 Wish you could decide in one line what your AI agent is allowed to do and what it must never touch? OpenCode's built-in tools plus permissions give you exactly that. 🏷️ Title: Built-in Tools + Permissions 🔗 URL: 📘 Overview OpenCode agents act on your codebase through "tools" such as file editing and shell execution. A rich set ships by default, and each tool can be governed by an allow / ask / deny policy. You get the safety-versus-convenience balance tuned entirely from config. ⚙️ How It Works The main built-in tools are: ・`bash`: run shell commands (git, npm, etc.) ・`edit`: modify existing files via exact string replacement ・`write`: create or overwrite files ・`read`: read files, with optional line ranges ・`grep`: regex search across files ・`glob`: find files by patterns like `**/*.js` ・`webfetch` / `websearch`: fetch and search the web ・helpers like `lsp`, `apply_patch`, `skill`, `todowrite`, `question` Permissions are set in the `permission` field with three states: `allow` (run freely), `ask` (confirm each time), `deny` (forbidden). Note that the `edit` permission governs `edit`, `write`, and `apply_patch` together. 🛠️ Practical Usage In `opencode.json`, you can forbid edits, confirm every bash call, and allow web fetches freely — set `"edit": "deny"`, `"bash": "ask"`, and `"webfetch": "allow"` under the `permission` block. Tools coming from MCP servers can be controlled with wildcards. Writing `"mymcp_*": "ask"` requires confirmation for every tool from that server. 💡 Use Cases On a production-adjacent repo, set `edit` to `deny` and `bash` to `ask` so the agent can plan and investigate but cannot rewrite code or run destructive commands on its own. On a throwaway experiment branch, allow everything to move fast. Switching between the two is just a config change. ⚠️ Caveats By default all tools are allowed, so nothing is restricted until you explicitly narrow it. The `lsp` tool needs `OPENCODE_EXPERIMENTAL_LSP_TOOL=true`, and `websearch` (powered by Exa) needs `OPENCODE_ENABLE_EXA=1`. It is easy to forget that the `edit` permission also covers write and apply_patch. #OpenCode# #AIAgents#

【TradFi Daily | 1-Minute Stock Spotlight】 Today's pick: $TQQQ Triple the daily move of the Nasdaq-100. 🚀 ✅ Delivers 3× the daily return of the Nasdaq-100 Index ✅ Exposure to tech leaders like Apple, Microsoft, NVIDIA, and Amazon ✅ Over $30B in assets under management ✅ Rebalanced daily for short-term trading When Big Tech rallies,TQQQ amplifies the momentum. 📈 Trade $TQQQ on Websea 👇

JSEF v1.10.0 is out 🎉 A model-agnostic LLM eval toolchain + 20+ new hard vulnerability samples. ✅ Swap any OpenAI/Anthropic-compatible model (GLM-5.3, Mythos, GPT…) ✅ One-click cross-model leaderboard (Recall/Precision/F1/accuracy) ✅ DeepSwe-style trials stability (Pass@1, N-run) ✅ Multi-state gate, JWT 4-step chain, async taint chain Star → #LLMSecurity# #AppSec# #WebSecurity# #JavaSecurity# #VulnerabilityDetection# #AISecurity#