[Cloudflare uses Streamable HTTP](https://github.com/cloudflare/mcp-server-cloudflare)reti

Monday, 21 September 2026

Cloudflare uses Streamable HTTP moves MCP servers to Streamable HTTP and returns 410 for legacy SSE GET. OpenAI tunnels private MCP connects private MCP servers without public exposure.

Lawsuit alleges AI slowdown and [Anthropic picks Accenture evalu](https://www.reddit.com/r

🛠️ZCode opens its coding work

GitHub·★ 3.1k·788 forks

ZCode's repository includes desktop, web, CLI, backend, shared UI and agent runtime source for its coding workbench.

Top Voted

TypeScriptAgentsElectron

⚖️Lawsuit alleges AI slowdown

Reddit (/LocalLLaMA)

A Reddit title reports allegations that Anthropic, OpenAI, SpaceXAI and Google agreed illegally to slow AI progress; details are not supplied.

Deep DiveAnthropicOpenAIGoogle

🧠Route agent decisions with Jev

GitHub·★ 288·24 forks

Hermes Jev Skills applies a typed decision model to model routing, memory, compaction, skill selection and browser or computer actions.

Deep Dive

PythonAgentsSecurity

🐢Tao calls for slower AI

Reddit (/OpenAI)

This Reddit item quotes mathematician Terence Tao arguing that AI development is moving too quickly, without supplying additional source context.

Lively ThreadAI SafetyRegulation

📚Browse source-backed Jev tools

GitHub·★ 277·47 forks

This curated list catalogues source-backed Jev projects for typed probabilistic classification, routing, scoring, ranking and guardrails.

TrendingAgentsEvalsOpen Source

🧩Share projects built withClaude

Reddit (/ClaudeAI)

This Reddit post asks people to share their proudest side projects made with Claude, but the supplied item contains no project examples.

DivisiveClaudeCode GenDev Tools

💻dscode adds agent sessions

GitHub·★ 201

dscode is a DeepSeek coding-agent harness with persistent shells, Ultra subagents, auto approval, Chrome MCP and session telemetry.

Bigger Picture

Agent session controls

Test persistent shells and subagent delegation with approval policies; inspect session telemetry before…

Trending

JavaScriptDeepSeekAgents

📰Trump faces mockery over AI

Reddit (/OpenAI)

This Reddit title says Trump was mocked over an effort to rename AI as 'superior' or 'extreme' intelligence, with no further evidence supplied.

AI SafetyRegulation

Learn/Multiple Mentions

What does inference mean in these tools?

Inference is a trained model producing outputs from input, and it's central to llm-visuals, Von and laya.cpp. Devs tune speed and cost.

LatencyThroughput

Yesterday's Sentiment/Cautiously practical

Concrete migrations, unvalidated claims

Cloudflare uses Streamable HTTP provides a concrete client migration, while llama.cpp PR targets sparse Q4 and Lawsuit alleges AI slowdown require validation.

⚡Von offers local decisions

GitHub·★ 243·22 forks

Von is described as an open-source, non-autoregressive local decision model and a drop-in alternative to TypeSafe Jev.

Trending

PythonInferenceLocal AI

🧪Anthropic picks Accenture evalu

Reddit (/ClaudeAI)

The title says Anthropic selected Accenture as its first embedded evaluator to help implement a slowdown proposal, without further details.

AnthropicEvalsAI Safety

🌐Use Stagehand for web tasks

GitHub·★ 24.7k·1.7k forks

Stagehand is a TypeScript SDK for extracting data from websites and interacting with them, with integrations including Claude Code and Codex.

Trending

TypeScriptAgentsComputer Vision

⚙️llama.cpp PR targets sparse Q4

Reddit (/LocalLLaMA)

A linked llama.cpp pull request proposes CUDA sparse Flash Attention support for Qwen4, but the supplied item gives no validation details.

C++CUDAInferencePerformance

🎨OpenCreator unifies local agent

GitHub·★ 12k·1.2k forks

OpenCreator is a local workspace using Codex CLI for creator workflows spanning video, images, voice, avatars, translation and editing.

Trending

TypeScriptOpenAICode Gen

💬Compare different LLM prompts

Reddit (/LocalLLaMA)

This Reddit post discusses how differently people prompt LLMs, but the supplied item gives no examples, findings or implementation guidance.

Most DiscussedLLM OpsEvalsResearch

📬Build an email agent stepwise

GitHub·★ 2.3k·443 forks

This guide builds an email agent from basics through evaluation, Gmail integration, human approval, memory and deployment with LangGraph.

Trending

PythonLangChainAgents

⚡Explore laya.cpp decisions

Reddit (/LocalLLaMA)

The Reddit title presents laya.cpp as optimised for near-instant decision making, but the supplied item contains no implementation evidence.

C++InferencePerformanceLocal AI

📊Monitor local LLM inference

GitHub·★ 98·14 forks

llm-visuals displays local server counters and GPU telemetry for throughput, TTFT, context, speculative decoding and MoE routing.

Bigger Picture

Inference telemetry

Use dashboard to correlate TTFT, throughput, context and speculative-decoding counters with NVIDIA or…

Trending

RustInferenceObservability

🔓ZCode becomes open source

Reddit (/LocalLLaMA)

This Reddit item reports that ZCode is now open source; the supplied repository separately describes desktop, web, CLI and runtime source.

TypeScriptOpen SourceAgents

🔗Share actions between agents

GitHub·★ 5.3k·497 forks

Agent-Native is a TypeScript framework where agents and UIs share validated actions, permissions, implementations, data and application state.

Trending

TypeScriptAgentsMCP

🕹️Qwen3.8 makes an arcade game

Reddit (/LocalLLaMA)

The Reddit title describes a Cosmic Arcade one-shot game made with Qwen3.8-Flash-Next, but supplies no implementation or evaluation details.

QwenCode GenOpen Source

🔬Review Jev papers and tests

GitHub·★ 114·6 forks

The Jev gallery collects papers, open reproductions and independent evaluations related to System One models and calibrated probabilities.

Trending

JavaScriptResearchEvals

🧰Review a two-year coding stack

Reddit (/AI_Agents)

This Reddit item reports a practitioner's day-to-day stack for AI-assisted, agent-oriented coding, but supplies no technical details.

AgentsDev ToolsCode Gen

🔧Cloudflare uses Streamable HTTP

GitHub·★ 4.3k·528 forks

Cloudflare's MCP servers use stateless Streamable HTTP at /mcp and /sse, while legacy SSE GET requests receive 410 Gone responses instead.

Bigger Picture

MCP migration test

Move MCP clients to Streamable HTTP now: legacy SSE GET calls return 410. Retain the [existing URL](https://github.com/cloudflare/mcp-server-clo.

Under The Radar

TypeScriptCloudflareMCP

🔒OpenAI tunnels private MCP

GitHub·★ 452·91 forks

OpenAI's customer-run client connects private or localhost MCP servers to ChatGPT, Codex, and APIs without directly exposing them publicly.

Bigger Picture

Private MCP tunnel checks

Run the tunnel client beside a private MCP server; probe /healthz, /readyz and /metrics before…

Trending

GoOpenAIMCP

Learn/Core Concept

How does speculative decoding speed inference?

Speculative decoding uses a smaller draft model to suggest tokens, then a larger model verifies them in batches. It can cut latency without changing output quality, as llm-visuals helps measure.

InferenceLatency

Read online