BlakeHung

[BlakeHung @mars7823482]

@sunsetsyntax good catch — the allowlist only covers the LLM side. a true air-gap means turning off Zed's own extension/update checks too. worth a note in the docs, thanks.

[BlakeHung @mars7823482]

Run local models in Zed with a hardware air-gap. 🔒

ACP is the agent-editor standard — wiring it to offline Ollama/llama.cpp was the missing piece. acp-bridge fills it: one binary, v1+v2 wire, outbound HTTP only via explicit allowlist.

https://t.co/loG9ZMc1Ht #ZedEditor #ACP

[BlakeHung @mars7823482]

@mustafa_homelab neither really — it's backend-agnostic. acp-bridge just speaks OpenAI-compatible `/v1`, so llama.cpp server, ollama, vllm all work. for the Strata post it was Strata's own engine at `127.0.0.1:8080/v1`.

[BlakeHung @mars7823482]

Local LLM in your editor: no cloud, no runner, no telemetry — and it speaks ACP natively.

Why and how I built acp-bridge in Rust: 🧵

[BlakeHung @mars7823482]

@mitansh_j07 thanks — that combo is the point: agent in the editor, model on your own box, nothing phones home. took it a step further and verified it against a 125B local model end to end: https://t.co/ZOC1PJ4zdL

[BlakeHung @mars7823482]

Design notes + the 0.9.2 tool-call recovery story:
https://t.co/loG9ZMc1Ht
Source & releases: https://t.co/n6jHfP8mVc
cargo install acp-bridge

[BlakeHung @mars7823482]

The constraints that shaped it:
1/ One ~5MB static binary
2/ ACP v1 + v2 dual wire
3/ Tools sandboxed to the session dir
4/ Outbound HTTP off by default
5/ 185 tests incl. e2e vs Zed / Codex CLI

[Pahud Hsieh @pahudnet]

The Agenda for tomorrow community meetup is out. Don't miss it! We'll have community Heroes @365_step_tech sharing level up your CDK DX and @fantasticHsieh on CDK using @deepseek_ai Harness. We'll cover our upcoming CDK PR Review Agent and the whole 2026Q3 recap for you! See you tomorrow!

[BlakeHung @mars7823482]

@openabdev 🥰🥰🥰

[BlakeHung @mars7823482]

Full write-up (14 controls + 5 department examples): https://t.co/1i5xrpKtsr

[BlakeHung @mars7823482]

Full write-up (14 controls + 5 department examples): https://t.co/1i5xrpKtsr

[BlakeHung @mars7823482]

I think frontend's next role is Frontend Deployment Engineer. IEEE Software 2026 (NSF-funded, 3 institutions): LLM → 100% same answer Humans → 8 choices, 21.7% said "need more info" Config = the boundary. Humans decide what's in it.

#FrontendDeploymentEngineer #OpenAB

[BlakeHung @mars7823482]

I think frontend's next role is Frontend Deployment Engineer. IEEE Software 2026 (NSF-funded, 3 institutions): LLM → 100% same answer Humans → 8 choices, 21.7% said "need more info" Config = the boundary. Humans decide what's in it.

#FrontendDeploymentEngineer #OpenAB

[Pahud Hsieh @pahudnet]

這是真的，很多時候特別是代碼場景是多個Agent 來自各家性價比最好的Model一起collab，互相糾錯，而不是花巨大的費用一家頂級model one-shot 完成。 model跟agent的多樣性與互相溝通與糾錯能力會是一個不可忽視的亮點。

[painn @painn_x]

Found a way to run GLM-5.2 for FREE with no daily limits through Cloudflare Workers AI 😳

No credit card. No monthly bills. Just one API token and you get solid usage.

What GLM-5.2 on Cloudflare unlocks:
-Strong agentic coding performance
-262k context window
-Tool use and function calling
-OpenAI-compatible endpoint
-Works with OpenCode, Cursor, Aider, Hermes, and more

What this replaces:
-Paid frontier model usage for daily coding agents
-Expensive API costs just to test workflows

How to grab it (5 min):

1. Create a free Cloudflare account

2. Go to Workers AI and copy your Account ID

3. Create an API Token

4. Add the provider in OpenCode with this base URL:
https://t.co/I06wzl3TnF

5. Select model: @cf/zai-org/glm-5.2

6. Start running agents

Works in OpenCode, Cursor, Aider, Claude Code, Hermes Agent, and any OpenAI-compatible tool.

Your buddy pays $20–200/mo for coding agents. You get strong performance for $0.

Bookmark this before Cloudflare changes the free tier.

[OpenAB - Open Agent Broker @openabdev]

Love this — we took it further.

OpenAB = tag a whole team. Claude, GPT, Grok & Gemini join one channel, take tasks, even deliberate together.

Open source. Discord, Telegram, WeCom, Feishu, LINE, Teams, GoogleChat — and Slack too. Not just Slack.

🔗 https://t.co/1eGaBMk9WL

[Pahud Hsieh @pahudnet]

AIコーディングエージェントはCDKを書くのは得意。でも、そのコンストラクトが実際にどんなインフラを作るかは見えていない。

CDK LSPサーバーのRFCを公開しました。エージェントにも人間にも、編集するたびにコンストラクト→リソースのマッピングとバリデーション結果を提供します。デプロイ不要。

さらに cdk explore — CDKアプリ全体をブラウザで視覚的にナビゲートできる3パネルUI。

あなたにとって一番大事な機能は？フィードバックお待ちしています 👇

https://t.co/eTaSjQHEvn

@365_step_tech
@nixieminton
@mazyu36
@jawsug_cdk
@WinterYukky

[BlakeHung @mars7823482]

https://t.co/jZCE3AP32i

[BlakeHung @mars7823482]

Enterprise AI gateway: feature list ≠ architecture.

4 principles for sensitive-data infra: • Control by Config • Security by Architecture (no HTTP) • Accountability by Default • Portability by Design (ACP)

#EnterpriseAI #AIGateway #AgenticAI #CyberSecurity

[BlakeHung @mars7823482]

Enterprise AI gateway: feature list ≠ architecture.

4 principles for sensitive-data infra: • Control by Config • Security by Architecture (no HTTP) • Accountability by Default • Portability by Design (ACP)

#EnterpriseAI #AIGateway #AgenticAI #CyberSecurity

[OpenAB - Open Agent Broker @openabdev]

🚀 OpenAB 0.8.4 — Native Rust coding agent drops! ~7MB/session, 28x lighter than Pi. OpenShell sandbox, lifecycle hooks, new backends.

Lightweight. Secure. Cloud-native.

[BlakeHung @mars7823482]

-> https://t.co/X0KIBuBt9G

[BlakeHung @mars7823482]

At Computex 2026, Jensen framed enterprise AI's real blocker as security review, not model capability. NVIDIA's answer: OpenShell — sandboxed agent runtime.

I'd written the same as a 3-layer trust architecture before the keynote. Same conclusion, different frame.

[BlakeHung @mars7823482]

At Computex 2026, Jensen framed enterprise AI's real blocker as security review, not model capability. NVIDIA's answer: OpenShell — sandboxed agent runtime.

I'd written the same as a 3-layer trust architecture before the keynote. Same conclusion, different frame.