⚡Trending:
H
Hugging Face Daily Papers@HuggingFace·1h ago
🔥 Trending

CodeGraph: First Open-Taxonomy Code Knowledge Graph Spans 158M Nodes and 1B Edges Across 167M Files

Developed by researchers from Software Heritage and European research universities, CodeGraph introduces the first large-scale open-taxonomy knowledge graph for source code. Scaling across 167 million files in the Stack-Edu corpus, it extracts implicit engineering abstractions—algorithms, architectural paradigms, and design patterns—via a code-specialized LLM and a three-stage Wikidata entity-linking pipeline, materializing 158 million nodes and approximately 1 billion typed edges across 14 languages.

CodeGraph: First Open-Taxonomy Code Knowledge Graph Spans 158M Nodes and 1B Edges Across 167M Files
⚡ Key Takeaways
  • •Beyond Syntactic AST Parsing: Moves past token and syntax tree representations by explicitly mapping source code implementations to high-level engineering paradigms, design patterns, and algorithmic taxonomy.
  • •Massive 158M Node and 1B Edge Scale: Built across 167 million files in Stack-Edu across 14 languages, capturing 145 million file nodes, 63,000 extracted concept entities, and 19,800 grounded Wikidata identifiers.
  • •Three-Stage Linking Pipeline: Combines deterministic SPARQL entity linking, a Deep Research Agent for residual long-tail disambiguation, and hierarchy rollup for parent closures, calibrated via human gold sets and LLM judges.
Read details→
H
Hugging Face Daily Papers@HuggingFace·1h ago
🔥 Trending

TimeEvo: Failure-Driven Self-Evolution of Time Series Agents Outperforms Human-Curated Toolkits

Researchers from UIC exposed two structural flaws in autonomous agents reliant on human-curated toolkits: expert-selected tools frequently degrade anomaly detection across all backbones, while unguided self-revision silently breaks 38% of previously correct responses. They introduce TimeEvo, an architecture that clusters runtime failures into capability gaps, synthesizes evidence-only Python tools to address them, and admits candidate tools via paired verification gates. Starting from an empty toolkit, TimeEvo achieves consistent accuracy gains across all ten benchmarks.

TimeEvo: Failure-Driven Self-Evolution of Time Series Agents Outperforms Human-Curated Toolkits
⚡ Key Takeaways
  • •Exposing Tool Misalignment & Silent Harm: Demonstrates that static expert toolsets paradoxically degrade anomaly accuracy across all LLM backbones, while generic self-revision silently degrades 38% of previously correct answers.
  • •Failure-Driven Evolution Pipeline: Systematically clusters execution failures into capability gaps, synthesizes dedicated Python evidence tools, and validates candidates through paired admission gates.
  • •Zero-Bootstrap Superiority & Cross-Model Transfer: Starting from a completely empty library, TimeEvo boosts performance across ten benchmarks and three backbones; tools evolved on cheap models reliably transfer gains to frontier foundation models.
Read details→
H
Hugging Face Daily Papers@HuggingFace·1h ago
🔥 Trending

Disaggregated Quantization: Specializing LLM Prefill and Decode Formats Cuts TTFT by 1.78x

Addressing the fundamental tension where prefill is compute-bound while decode is memory-bandwidth-bound, researchers from ISTA and Neural Magic introduced Disaggregated Quantization (DQ). By assigning compute-native NVFP4 weights to prefill and compact 2-3 bit weights to decode, paired with Offloaded Disaggregated Prefill (ODP) that streams prefill weights from SSD, DQ cuts Time-To-First-Token (TTFT) by 1.78x at 8K context and lifts 1-bit decode accuracy by 32.5 points on MMLU-Pro.

Disaggregated Quantization: Specializing LLM Prefill and Decode Formats Cuts TTFT by 1.78x
⚡ Key Takeaways
  • •Stage-Specialized Formats: Eliminates uniform end-to-end quantization by applying compute-native NVFP4 for compute-bound prefill and compact 2-3 bit representations for bandwidth-bound decode token generation.
  • •Zero-Footprint SSD Streaming: Offloaded Disaggregated Prefill (ODP) dynamically streams specialized prefill weights from high-speed SSDs during prompt ingestion, completely amortizing I/O latency across sequence lengths.
  • •1.78x TTFT Speedup and 2.8T Scale: Delivers a 1.78x TTFT speedup on Qwen 27B at 8K prompt length in llama.cpp, raises 1-bit decode accuracy by 32.5 points on MMLU-Pro, and validates up to 2.8T parameter scales under vLLM disaggregation.
Read details→
m
matt palmer@mattyp·1h ago
🚀 Release

Cursor Self-Hosted: one `agent worker start` turns a Mac Mini into a remote worker; `--computer-use` drives the desktop

Cursor engineer @mattyp shows a Mac Mini becoming a Cloud Agent worker via `agent worker start`; `--computer-use` enables remote click/type/screenshot through Cursor Computer Use. Official Self-Hosted Machines keep inference in Cursor cloud and tool execution on your hardware over outbound HTTPS (My Machines + Team Pools).

Cursor Self-Hosted: one `agent worker start` turns a Mac Mini into a remote worker; `--computer-use` drives the desktop
⚡ Key Takeaways
  • •Product path: Self-Hosted Machines blog + docs — tools on your hardware, loop in Cursor cloud
  • •Personal quickstart (My Machines): install CLI → agent login → agent worker start (outbound HTTPS only)
  • •Desktop control: agent worker --computer-use start; grant Accessibility + Screen Recording to Cursor Computer Use on macOS
  • •Scale: 200 workers/user, 1000/team; Team Pools need Enterprise + service-account API key
  • •Verify with agent worker debug + screenshot task; route via worker= from Slack/GitHub/Linear
Read details→
ADSponsored
A
Anthropic@AnthropicAI·2h ago
🚀 Release

Anthropic launches Claude Sonnet 5.5: Terminal-Bench 70.6%, 30%+ faster output, same $2/$10 price with lower task cost

Anthropic released Claude Sonnet 5.5 on 2026-09-28 (API id `claude-sonnet-5-5`), the second Claude 5.5-family model. Vs Sonnet 5: Terminal-Bench 4.0 70.6% (from 10.3%), CursorBench 4.0 55.5%, GDPval-AA near Opus 5.5; 30%+ faster output and up to ~30% lower task cost at the same $2/$10 pricing. Native 1M context; adaptive thinking on by default — use `between_tools` to disable up-front thinking.

Anthropic launches Claude Sonnet 5.5: Terminal-Bench 70.6%, 30%+ faster output, same $2/$10 price with lower task cost
⚡ Key Takeaways
  • •Shipped 2026-09-28 — model id claude-sonnet-5-5 (announcement)
  • •Terminal-Bench 4.0 70.6% (vs Sonnet 5 10.3%); CursorBench 4.0 55.5%; GDPval-AA 1844
  • •Same $2/$10 pricing; ~30%+ faster output; up to ~30% lower task cost via fewer tokens
  • •1M context / 128K max out; adaptive thinking default; between_tools turns off up-front thinking
  • •Migrate via docs; anthropic-sdk-python ≥1.9.0; Claude Code ≥2.1.284 sonnet alias → 5.5
Read details→
I
InternLM / Shanghai AI Lab@InternLM·4h ago
🚀 Release

LMDeploy 0.18.0: Qwen3.5 dflash, SM90 quantized GEMMs, Ascend GLM-5.2; TurboMind W4A16 + piecewise CUDA Graph

Shanghai AI Lab’s LMDeploy shipped v0.18.0 on 2026-09-28: input logprobs, expanded SM90 quantized GEMMs with unified TurboMind linear paths, Ascend GLM-5.2, Qwen3.5 dflash, and MoE shared-expert/FFN sharding. PyTorch path prototypes TurboMind W4A16 (AWQ), piecewise CUDA Graph prefill, XTuner TileLang sparse MLA, checkpoint-engine weight updates, and request-only KV cache metrics.

LMDeploy 0.18.0: Qwen3.5 dflash, SM90 quantized GEMMs, Ascend GLM-5.2; TurboMind W4A16 + piecewise CUDA Graph
⚡ Key Takeaways
  • •Shipped v0.18.0 — Qwen3.5 dflash, Ascend GLM-5.2, SM90 quantized GEMMs
  • •TurboMind W4A16 AWQ prototype + piecewise CUDA Graph prefill + TileLang sparse MLA
  • •checkpoint-engine weight updates + request-only KV cache metrics + input logprobs
  • •Upgrade: pip install -U "lmdeploy>=0.18.0"
Read details→
M
Model Context Protocol@modelcontextprotocol·4h ago
🛠️ Tooling

MCP TypeScript SDK 2.2.0: M2M OAuth expectedIssuer, auto-paging list*, CJS jose types + Client.listen hang fixes

MCP’s TypeScript monorepo shipped v2.2.0 on 2026-09-28: client/server/core/server-legacy/codemod aligned at 2.2.0. M2M OAuth providers should pass `expectedIssuer`; `fetchToken()` throws `AuthorizationServerMismatchError` on issuer mismatch. Cursor-less listTools/listPrompts/listResources/listResourceTemplates now follow `nextCursor` (capped by `listMaxPages`). Fixes the 2.1.0 CJS jose types regression and Client.listen() unhandled rejection / send hang.

MCP TypeScript SDK 2.2.0: M2M OAuth expectedIssuer, auto-paging list*, CJS jose types + Client.listen hang fixes
⚡ Key Takeaways
  • •Shipped v2.2.0 — client/server/core/server-legacy/codemod aligned; node/express/hono/fastify unchanged
  • •M2M OAuth: pass expectedIssuer; fetchToken throws AuthorizationServerMismatchError on mismatch
  • •list* without cursor auto-follows nextCursor (listMaxPages cap)
  • •Fixes CJS jose types regression and Client.listen() rejection/hang
  • •Install: npm i @modelcontextprotocol/[email protected]
Read details→
A
Anthropic@AnthropicAI·4h ago
🛠️ Tooling

Anthropic Python SDK 1.9.0: claude-sonnet-5-5, between_tools thinking, cache diagnostics GA, tools while reply streams

Anthropic shipped Python SDK v1.9.0 on 2026-09-28: adds claude-sonnet-5-5, between_tools thinking, GA cache diagnostics on Message/MessageCreateParams, and optional tool execution while the reply streams. Managed Agents event filters and workspace rate-limit include_inherited/source land too, plus unnamed upload and stream()/parse() diagnostics fixes.

Anthropic Python SDK 1.9.0: claude-sonnet-5-5, between_tools thinking, cache diagnostics GA, tools while reply streams
⚡ Key Takeaways
  • •Shipped v1.9.0 — 31 commits vs v1.8.0; claude-sonnet-5-5 + between_tools thinking
  • •Cache diagnostics GA on Message / MessageCreateParams; stream()/parse() accept diagnostics
  • •Optional tool execution while the assistant reply streams
  • •Upgrade: pip install -U "anthropic>=1.9.0" (Python 3.10+)
Read details→
ADSponsored
L
LangChain4j@langchain4j·9h ago
🛠️ Tooling

LangChain4j 1.20.2: A2A client tenant propagation—auto-attach tenant on messages without exposing it as an LLM tool arg

LangChain4j 1.20.2 (2026-09-28) focuses on Agent-to-Agent (A2A) client fixes: `tenant()` on `A2AClientInstance`/`A2AClientBuilder`, `@A2AClientAgent(tenant=…)` auto-attaches tenant on every `MessageSendParams` (or parses it from `/.well-known/{tenant}/agent-card.json`), and filters tenant method args so they are not exposed as LLM tool parameters. AgenticScope/ResultWithAgenticScope toString guards against circular refs.

LangChain4j 1.20.2: A2A client tenant propagation—auto-attach tenant on messages without exposing it as an LLM tool arg
⚡ Key Takeaways
  • •Shipped 1.20.2 — A2A tenant auto-propagation via annotation/builder or Agent Card URL
  • •Tenant args filtered from LLM tool surface; attached on MessageSendParams
  • •AgenticScope toString circular-ref guard
  • •Docs: a2a-protocol.org + docs.langchain4j.info/tutorials/agents
Read details→
U
Unsloth@unslothai·9h ago
🛠️ Tooling

Unsloth v0.1.900-beta: local Laya Decision models + Skills editor; ~4.5× LTX-2.3 clips and up to 6.3× faster VAE decode

Unsloth shipped v0.1.900-beta on 2026-09-28 (PyPI unsloth 2026.9.12): Desktop runs Laya Decision models locally via a Jev-compatible `/v1/systemone` endpoint, adds a Skills CRUD editor and Library viewers for PDF/Office, and improves Apple Silicon with batched serving, structured outputs, and TurboQuant KV. Media pipelines claim ~4.5× LTX-2.3 clips, 1.7–6.3× VAE decode, and up to ~1 minute faster MiniMax-H3 first render with 25–29 GiB lower peak VRAM.

Unsloth v0.1.900-beta: local Laya Decision models + Skills editor; ~4.5× LTX-2.3 clips and up to 6.3× faster VAE decode
⚡ Key Takeaways
  • •Shipped v0.1.900-beta — local Laya Decision API + Skills editor + Library viewers
  • •LTX-2.3 ~4.5×; VAE decode 1.7–6.3×; MiniMax-H3 first render up to ~1 min faster, −25–29 GiB peak VRAM
  • •Apple Silicon: batched serving, structured outputs, TurboQuant KV; ModelScope downloads
  • •Docs: docs.unsloth.ai
Read details→
O
OpenAI@OpenAI·9h ago
🛠️ Tooling

OpenAI Python SDK 3.20.0: Agents credential/session options, Cyber access programs on Responses, opt-in WebSocket text/tool snapshots

OpenAI shipped Python SDK v3.20.0 on 2026-09-28: Agents credentials gain metadata and update-without-replace auth, sessions add ultrafast tier plus misalignment_policy_violation; Responses can select Cyber access programs (standard/daybreak_blue/daybreak_red); and an opt-in ResponsesWebSocketAccumulator yields immutable incremental text/tool snapshots. Live/Realtime WebSocket query/queue/TLS retry fixes land in the same cut.

OpenAI Python SDK 3.20.0: Agents credential/session options, Cyber access programs on Responses, opt-in WebSocket text/tool snapshots
⚡ Key Takeaways
  • •Shipped v3.20.0 — Agents credential metadata/session ultrafast + Cyber access programs
  • •Opt-in ResponsesWebSocketAccumulator for incremental text/tool snapshots (190+ WS tests)
  • •Live/Realtime WebSocket query/queue/TLS retry hardening
  • •Upgrade: pip install -U openai
Read details→
F
Firecrawl@firecrawl·13h ago
🛠️ Tooling

Firecrawl SDKs ship agent_hints: Python 4.45.0 / JS 4.42.0 preserve deterministic next-step guidance for agents

Firecrawl published firecrawl-py 4.45.0 and @mendable/firecrawl-js 4.42.0 on 2026-09-28, shipping `agent_hints` preservation from [#4641](https://github.com/firecrawl/firecrawl/pull/4641): with `X-Firecrawl-Agent-Hints: true`, Search/Scrape/Parse/Map responses can include deterministic next-step suggestions. Also includes Hangar browser/interact migration and `exchange.onTermsRequired` fields.

Firecrawl SDKs ship agent_hints: Python 4.45.0 / JS 4.42.0 preserve deterministic next-step guidance for agents
⚡ Key Takeaways
  • •Shipped firecrawl-py 4.45.0 + JS 4.42.0 with agent_hints preservation
  • •Opt-in via X-Firecrawl-Agent-Hints: true (off by default)
  • •Deterministic next-step rules for Search/Scrape/Parse/Map
  • •MCP pin still on 4.40.0 — bump needed for full hint survival
Read details→
O
OpenCode@anomalyco·13h ago
🛠️ Tooling

OpenCode 1.18.33: Cloudflare AI Gateway honors timeouts, MCP browser launch failures surface, debug config redacts credentials

OpenCode v1.18.33 (2026-09-28) makes Cloudflare AI Gateway models honor provider response/stream timeouts, reports MCP browser launcher exits, redacts credentials in debug config output, and aligns Gemini thinking defaults/effort options with supported controls across model generations.

OpenCode 1.18.33: Cloudflare AI Gateway honors timeouts, MCP browser launch failures surface, debug config redacts credentials
⚡ Key Takeaways
  • •Shipped v1.18.33 — CF AI Gateway timeouts honored
  • •MCP browser launcher immediate-exit failures reported
  • •Debug config redacts credentials/sensitive headers
  • •Gemini thinking defaults/effort aligned across generations
Read details→
D
Docker@docker·13h ago
🛠️ Tooling

Docker cagent 1.145.0: ACP request traces via W3C traceparent/tracestate, server spans on 13 handlers, lean TUI scrollback fix

Docker shipped cagent (docker-agent) v1.145.0 on 2026-09-28: ACP requests now propagate W3C `traceparent`/`tracestate` with server spans for each of the 13 implemented protocol handlers, plus new lint cops and a lean-TUI fix that preserves scrollback during partial streaming tool calls.

Docker cagent 1.145.0: ACP request traces via W3C traceparent/tracestate, server spans on 13 handlers, lean TUI scrollback fix
⚡ Key Takeaways
  • •Shipped v1.145.0 — ACP W3C traceparent/tracestate + 13 handler spans
  • •Lean TUI keeps scrollback during partial streaming tool calls
  • •New FieldsSeqLookup / SlicesConcat lint cops
  • •Docs: docker.github.io/docker-agent
Read details→
H
Hugging Face Daily Papers@HuggingFace·15h ago
🔥 Trending

Continuous Depth Batching (CDB): Unlocking Depth-Adaptive Inference for Looped Language Models with 99% Speedup Realization

While looped language models enable depth-adaptive inference by executing fewer layer repetitions on easy tokens and more on complex tokens, variable loop counts break conventional batching engines like vLLM. Researchers from TUM and Imperial College introduced Continuous Depth Batching (CDB). By dynamically reorganizing batches between loop iterations, orchestrating looped KV-caches, and asynchronously predicting token exits, CDB achieves up to 99% of the theoretical maximum inference speedup across Ouro 1.4B and Huginn 3.5B architectures.

Continuous Depth Batching (CDB): Unlocking Depth-Adaptive Inference for Looped Language Models with 99% Speedup Realization
⚡ Key Takeaways
  • •Resolving the Looped Batching Impasse: Depth-adaptive inference allows tokens to exit recurrent transformer layers early, but differing loop counts prevent uniform forward passes in engines like vLLM; CDB introduces inter-step batch reformation.
  • •Asynchronous Exit Prediction & Looped KV Cache: Integrates a dynamic scheduler between recurrent blocks that manages recurrent KV-cache states and asynchronously prepares future batches based on exit-likelihood estimates.
  • •99% Theoretical Limit Realized: Validated on Ouro 1.4B and Huginn 3.5B, CDB achieves up to 99% of the upper-bound theoretical speedup, clearing the path for production deployment of adaptive-depth architectures.
Read details→
H
Hugging Face Daily Papers@HuggingFace·16h ago
🔥 Trending

InternW0-Δ Open-Sourced: Unified World Action Model Pretrained on 20,000+ Hours of Open Robotics Data

Shanghai AI Lab and OpenDataLab open-sourced InternW0-Δ, a unified World Action Model (WAM) pretrained on over 20,000 hours of curated robotic demonstrations, UMI data, and egocentric videos. Built on a Mixture-of-Transformers (MoT) architecture that binds visual predictive dynamics with robot actions, it introduces 'Causal Imprint' to inject future representations directly into the action expert without requiring costly future-video rollouts during inference.

InternW0-Δ Open-Sourced: Unified World Action Model Pretrained on 20,000+ Hours of Open Robotics Data
⚡ Key Takeaways
  • •20,000+ Hours Open-Source Corpus: Unifies robot manipulation, UMI dexterous data, and egocentric human demonstrations into the largest publicly released robotic training corpus.
  • •Mixture-of-Transformers (MoT): Coordinates video dynamics experts and action generation experts under semantic guidance from a frozen VLM, with 4D spatial priors distilled during pretraining.
  • •Causal Imprint Without Future Rollout: Eliminates the critical latency bottleneck of traditional WAMs that require rendering future video frames at test time, providing predictive latent cues directly for real-time control.
  • •Complete Open Source: Training code, model weights, data processing pipelines, and datasets released at internrobotics.github.io/InternW0-Delta/.
Read details→
G
Google DeepMind@GoogleDeepMind·16h ago
📊 Benchmark

Google DeepMind & Kaggle Launch Game Arena: Dynamic Competitive Benchmarks for LLM Strategic Evaluation

Addressing the rapid saturation and benchmark overfitting of static LLM evaluations, Google DeepMind and Kaggle unveiled Game Arena. By orchestrating head-to-head live agent matchups with dynamic Elo rating evolution, Game Arena prevents performance saturation across three pilot environments: Chess (perfect information), Texas Hold'em Poker (imperfect information), and Werewolf (multiplayer social deduction and deception), systematically benchmarking long-horizon strategic reasoning.

Google DeepMind & Kaggle Launch Game Arena: Dynamic Competitive Benchmarks for LLM Strategic Evaluation
⚡ Key Takeaways
  • •Dynamic Head-to-Head Evaluation: Replaces static, easily saturated benchmarks with real-time model-versus-model matchups and dynamic Elo tracking that scales naturally as LLM reasoning capabilities evolve.
  • •Comprehensive Game Theory Spectrum: Spans perfect information (Chess), imperfect information with strategic betting (Poker), and multi-turn multi-agent social deduction and deception (Werewolf).
  • •Reproducible Sandboxed Infrastructure: Jointly open-sourced by DeepMind and Kaggle, providing standardized Python tournament APIs, replay verification pipelines, and sandboxed anti-cheat guardrails.
Read details→
G
Google@google·18h ago
🚀 Release

Google ADK Python 2.10.0: experimental skill lifecycle, MongoDB vector/hybrid search, eval duration/token metrics

Google shipped ADK Python v2.10.0 on 2026-09-25 (`google-adk` 2.10.0): experimental skill lifecycle (`ADK_ENABLE_SKILL_LIFECYCLE=1`) with ephemeral lifecycles and active skill limits; MongoDB toolset for vector/hybrid search; evaluation metrics for duration, tokens, and model-call counts; better OpenAI reasoning model parameter adaptation and reasoning-token reporting. BigQuery protected write mode and instruction templating tighten with breaking changes.

Google ADK Python 2.10.0: experimental skill lifecycle, MongoDB vector/hybrid search, eval duration/token metrics
⚡ Key Takeaways
  • •Shipped v2.10.0 — pip install google-adk==2.10.0; docs at adk.dev
  • •Experimental skill lifecycle via ADK_ENABLE_SKILL_LIFECYCLE=1
  • •MongoDB toolset for in-flow vector/hybrid search
  • •Eval metrics: duration, tokens, model-call counts
  • •Breaking: tighter BigQuery protected writes; OpenAIResponsesLlm uses effort not thinking_config; empty eval sets raise ValueError
Read details→
B
BerriAI@BerriAI·18h ago
🚀 Release

LiteLLM 1.103.0: MCP delegated OAuth admission, Fuse/Capability routers, prompt-cache cost prediction

BerriAI shipped LiteLLM v1.103.0 on 2026-09-28: delegated MCP OAuth now requires admission; Capability classifier + Fuse V2 routers land; the proxy can predict prompt-cache costs across deployments, bind JWT claims to agents, add `tpd_limit`, bulk user/team APIs, and `/nvidia_nim` passthrough. Also: Bedrock S3 managed file delete/list, Friendli price auto-sync, Responses↔Chat reasoning mapping, and cosign image verification docs.

LiteLLM 1.103.0: MCP delegated OAuth admission, Fuse/Capability routers, prompt-cache cost prediction
⚡ Key Takeaways
  • •Shipped v1.103.0 — pip install litellm==1.103.0; docs.litellm.ai
  • •MCP delegated OAuth requires admission; cosign-verify Docker images
  • •Capability + Fuse V2 routers; cross-deployment prompt-cache cost prediction
  • •JWT→agent binding, tpd_limit, bulk user/team management APIs
  • •Bedrock S3 file delete/list, Friendli price sync, Codex model picker sync from proxy
Read details→
O
OpenAI@openai·18h ago
🛠️ Tooling

OpenAI Codex 0.158.0: MCP pre-registered OAuth, bearer-secured exec-server, fullscreen TUI paste & transparent image gen

OpenAI shipped Codex rust-v0.158.0 on 2026-09-28 (`@openai/codex` 0.158.0): MCP servers with pre-registered OAuth client secrets (`codex mcp add --oauth-client-secret`); bearer-token auth for exec-server WebSockets; fullscreen TUI copy-on-select/right-click paste with Markdown-preserving copies; explicit transparent backgrounds for image gen/edit. Elevated-permission terminal approvals default on; Windows/Linux/macOS sandbox and approval-retry fixes land too.

OpenAI Codex 0.158.0: MCP pre-registered OAuth, bearer-secured exec-server, fullscreen TUI paste & transparent image gen
⚡ Key Takeaways
  • •Shipped rust-v0.158.0 / @openai/codex 0.158.0 — docs: CLI + MCP
  • •MCP pre-registered OAuth via codex mcp add --oauth-client-secret
  • •Bearer-secured exec-server WebSockets; elevated terminal approvals on by default
  • •Fullscreen TUI copy/paste keeps Markdown; Mermaid quoted labels/& supported
  • •Sandbox fixes across Windows 10, Linux nested writable roots, macOS path aliases
Read details→