AICoder

AI Coding Live News

First-hand AI coding updates from 𝕏/Twitter frontier labs and researchers, distilled by AI with entity benchmarks.

Explore live on 𝕏 ↗

AI Coding Live News

M
Mistral AI@MistralAI·
👀41
🔥85
View on 𝕏 ↗

Mistral Studio Adds Managed Agents and Apps: Durable Background Agents, Natural-Language Workflow Builder, HTTP Connectors and Online Evals

On November 5 Mistral expanded Mistral Studio. Managed Agents, built with the Agents SDK (mistralai-agents), run the same code in-process, self-hosted (agents serve) or Mistral-managed (agents deploy), with durable execution for hours or days, crash resume and human approvals. Mistral also shipped a Vibe-powered workflow builder, the mistral apps CLI, HTTP API connectors, webhooks and triggers, a managed search index, and offline and online evaluations, mostly in Public Preview.

Mistral Studio Adds Managed Agents and Apps: Durable Background Agents, Natural-Language Workflow Builder, HTTP Connectors and Online Evals

Key Takeaways

  • One agent codebase, three run modes: in-process, self-hosted agents serve, or managed agents deploy
  • Durable background execution for hours or days, resuming from the last finished step; Mistral stores sessions
  • Connectors now speak HTTP APIs with no MCP server needed; webhooks and triggers start workflows or agents
  • Managed agents emit agent, model and tool spans by default, and traces can become eval datasets in one click
  • Observability and evaluations are Enterprise-tier; most new features are Public Preview (vendor statement)
Read full story
Linked Models & Agents:
⚡ Mistral Large 4🤖 mistral-managed-agents🤖 mistral-vibe🌐 mistral
#Mistral#Mistral Studio#Managed Agents#Workflows#Evaluations#MCP
V
Vercelvercel·
👀84
🔥83
View on 𝕏 ↗

Vercel open-sources durable-harness: Claude Code, Codex, OpenCode and Pi agents resume from a durable log after crashes instead of starting over

On Oct 20, Vercel released durable-harness, an open-source package (Apache 2.0, npm 0.0.1-beta.6, marked experimental). It adds durability from the outside to existing coding-agent harnesses such as Pi, OpenCode, Codex and Claude Code: every model and tool call is logged before and after it runs in a Workflow SDK World, and if a function times out or crashes, a queue hands the turn to a new function that resumes from the log, so finished model calls are not billed twice. It runs on Vercel, on local files, or on self-hosted Postgres and other backends.

Vercel open-sources durable-harness: Claude Code, Codex, OpenCode and Pi agents resume from a durable log after crashes instead of starting over

Key Takeaways

  • Supports 5 harnesses: Pi, OpenCode, Codex, Claude Code and libfx's fx (eve and fx are already durable on their own)
  • Mechanism: an intent is recorded before each model or tool call and the outcome after it; each turn is backstopped by a queue message so a new function resumes from the log
  • Finished model calls are read back from the log, so Vercel says you never pay twice for the same call's tokens; only tools marked replay: 'safe' are rerun
  • Swappable backends: Vercel World, local files, Postgres, plus MongoDB, Redis and Turso; two environment variables move it to Postgres
  • Status: Apache 2.0, npm 0.0.1-beta.6, and the README marks it experimental with an API that may change
Read full story
Linked Models & Agents:
🤖 Claude Code🤖 Codex🤖 OpenCode🤖 Pi🤖 fx🤖 eve🌐 Vercel
#Durable execution#Coding agents#Workflow SDK#Open source#Agent infra
D
Denodeno_land·
👀129
🔥86
View on 𝕏 ↗

Deno team joins Cloudflare: Deno runtime gets one more year of fixes, Deno Deploy shuts down in six months, celld folds into workerd for self-hosted agent infra

On Oct 9 Ryan Dahl announced the whole Deno team is joining Cloudflare. Deno runtime gets one more year of monthly bugfix/security releases, then official development ends (it stays open source); Deno Deploy runs six more months before shutting down, with migration help to Workers; JSR continues. Focus shifts to the Workers/Durable Objects model and merging self-hosted celld into workerd, pitched explicitly at agent harnesses.

Deno team joins Cloudflare: Deno runtime gets one more year of fixes, Deno Deploy shuts down in six months, celld folds into workerd for self-hosted agent infra

Key Takeaways

  • Timeline: Deno runtime gets 12 months of monthly fixes then official development ends; Deno Deploy shuts down in 6 months with migration support for paying customers
  • JSR keeps running on Cloudflare infra; rusty_v8 stays supported and heads into workerd
  • celld (Apache-2.0, ~5.3k stars, v0.6.2 beta): a 58 MB static binary for self-hosted distributed Durable Objects backed by one S3-compatible bucket
  • Vendor-reported celld numbers: 0.47 MB RAM per resident cell, 2,500 cells per 8 GB node, 0.2/0.3 ms p50/p99 stateless requests, ~20 s failover with zero lost acknowledged writes
  • 1,234 points on the Hacker News front page; Deno's announcement post drew 6.3k likes
Read full story
Linked Models & Agents:
🌐 Deno🌐 Cloudflare
#Deno#Cloudflare#Workers#Durable Objects#Celld#Workerd
H
Hugging Face Daily Papers@HuggingFace·
👀89
🔥98

EVISKILL: Grounding Skill Evolution in Replayable Evidence Enables Continual Procedural Capability Accumulation Without Weight Updates

Continual skill evolution allows LLM agents to autonomously accumulate, refine, and reuse procedural knowledge across multi-turn interactions without updating model parameters. However, prevailing experience-driven reflection mechanisms suffer from two fundamental weaknesses: they decouple synthesized guidance from concrete behavioral traces, and they rely on monolithic global validation that blindly discards valid local corrections whenever an entire workflow fails. Researchers from Tsinghua University and partner institutions introduce EVISKILL (arXiv:2610.05030), an evidence-grounded framework anchoring procedural adaptation to deterministic execution traces. EVISKILL structures raw observations into Replayable Evidence Cards and synthesizes skill edits with explicit contextual lineage. A targeted replay engine verifies edits through isolated re-execution before cross-epoch refinement. Across three interactive benchmarks spanning six LLM backbones, EVISKILL consistently outperforms state-of-the-art reflective baselines, demonstrating durable procedural accumulation.

Key Takeaways

  • Tsinghua University introduces EVISKILL, anchoring continual procedural skill evolution in Replayable Evidence Cards without parameter updates
  • Decouples local edit verification from monolithic global validation, achieving a 3x increase in the retention of valid localized skills
  • Validated across 3 interactive benchmarks and 6 foundation LLMs, consistently outperforming standard reflection architectures
Read full story
Linked Models & Agents:
⚡ EVISKILL⚡ LLM-Agents⚡ Reflexion⚡ Voyager🤖 Skill-Evolution-Agent🤖 Replay-Validator🤖 Evidence-Curator🌐 Tsinghua University🌐 Hugging Face
#EVISKILL#Skill Evolution#Agent Memory#Replayable Evidence#Procedural Knowledge#Continual Learning
ADSponsored
T
T3 Code / pingdotgg@t3dotgg·
👀1.4k
🔥97

T3 Code Orchestrator V2 first nightly: cross-provider delegate_task, durable server queue; Pi 1.0 & OpenCode 2

pingdotgg/t3code shipped nightly v0.0.46-nightly.20261003.2610 at 2026-10-03T01:10:51Z — the first Orchestrator V2 build. It rebuilds turn lifecycle, queues/resume, subagent lineage, cross-provider threads, and history. Headline: delegate_task (spawn child agents on any provider/model and await results — the recommended multi-harness path, e.g. Claude plans / Codex implements), plus T3 Code MCP, ACP Registry, Cursor official SDK, restart-durable server queue, usage-limit auto-resume, scheduled tasks. At ingest: 24355★. Caveats: one-way DB copy from V1; mobile needs V2 beta; OpenCode 2.0.18+.

Key Takeaways

  • Release v0.0.46-nightly.20261003.2610 at 2026-10-03T01:10:51Z; site t3.codes
  • At ingest: 24355★; migration notes in issue #14871
  • V2: delegate_task, T3 Code MCP, ACP Registry, Cursor SDK, durable server queue, usage-limit auto-resume, scheduled tasks
  • New: Pi (incl. 1.0) + OpenCode 2 (≥2.0.18); mins Codex 0.159+ / Claude Code 2.1.280+ / Grok 1.0.13+ / Pi 0.80.5+
  • Caveats: one-way V1→V2 DB copy; V2 clients only on mobile beta; prefer delegate_task over lossy mid-thread provider switches
Read full story
Linked Models & Agents:
🤖 t3-code🤖 Claude Code (Anthropic)🤖 codex🤖 opencode🤖 pi🤖 cursor🤖 grok🤖 Google Antigravity (AGY)🌐 pingdotgg🌐 t3
#T3 Code#Orchestrator V2#delegate_task#Pi#OpenCode 2#Claude Code#Codex#Cursor SDK#ACP Registry#agent orchestration#Pingdotgg
E
Earendil@earendil·
👀148
🔥98

Earendil ships Pi 1.0: stable minimal agent harness with Codemode+MCP; ~111.8k★ / HN 1645 pts

On 2026-10-01 Earendil shipped Pi 1.0 (MIT): a hardened minimal coding-agent harness. Built-ins: Codemode (QuickJS sandbox for model-written JS calling tools / Jev / image models) + MCP, virtual-model routing, deferred tools, Anthropic cache warming, mid-conversation system edits. At ingest: 111793★; HN 1645 pts / 573 comments. Experimental Pi Durable shipped the same day.

Key Takeaways

  • Official post 2026-10-01; GitHub v1.0.0 at 2026-10-01T19:20:55Z; MIT
  • At ingest: 111793★; HN Pi 1.0 thread 1645 pts / 573 comments
  • 1.0: Codemode+MCP, virtual models, deferred tools, Anthropic cache warming, mid-convo system messages, fullscreen TUI; changelog ~5300→3300 prompt tokens (−40%) for GPT-5.6 with defaults+codemode
  • Not NVIDIA SoL-Pi (already in D1) — this is Earendil’s Pi harness
  • Install: curl https://pi.dev/install.sh | sh or npm @earendil-works/pi-coding-agent; docs pi.dev
Read full story
Linked Models & Agents:
⚡ jev⚡ GPT-6 Luna⚡ Claude Opus🤖 pi🤖 pi-durable🌐 earendil
#Pi#Pi 1.0#Earendil#agent harness#Codemode#MCP#virtual models#MIT#TUI#coding agent#Pi Durable
M
Mastramastra-ai·
👀53
🔥93

Mastra @mastra/core 1.72.0: live channel resolvers, lease-fenced background tasks, Microsoft Teams channel

Mastra shipped @mastra/[email protected] on 2026-09-30: live channel resolvers, lease-fenced background tasks, durable/evented recovery, aggregateTraces(), and @mastra/teams. Breaking: respondToToolApproval requires toolCallId.

Mastra @mastra/core 1.72.0: live channel resolvers, lease-fenced background tasks, Microsoft Teams channel

Key Takeaways

  • Release @mastra/[email protected] at 2026-09-30T10:31:27Z
  • Live channels() resolvers + resolveChannels() avoid redeploys
  • Lease-fenced background tasks with ownerId + expectedOwnerId writes
  • New @mastra/[email protected] and 25 Teams tools in Connect
  • Breaking: respondToToolApproval requires toolCallId
Read full story
Linked Models & Agents:
🤖 Mastra Agent🤖 EventedAgent🌐 Mastra🌐 Microsoft Teams
#Mastra#Agents#Teams#Channels#Leases#Observability#Durable#Release#Coding agent
B
ByteDance DeerFlow@bytedance·
👀130
🔥95

DeerFlow 2.1.0: 772-PR milestone — verifiable subagent delegation, pluggable memory, five remote sandboxes

ByteDance’s DeerFlow shipped v2.1.0 on 2026-09-24, closing the milestone with 772 merged PRs (~83k GitHub stars). Focus: trust and ops—verifiable subagent delegation, optional durable batch_task, pluggable memory (DeerMem / OpenViking / mem0 HTTP), plus E2B, BoxLite micro-VM, Tenki and OpenSandbox providers. Breakings: unconditional X-Trace-Id, reserved /mnt/skills, memory config under memory.backend_config, Docker bind defaults to 127.0.0.1.

DeerFlow 2.1.0: 772-PR milestone — verifiable subagent delegation, pluggable memory, five remote sandboxes

Key Takeaways

  • Shipped: v2.1.0 — 772 PRs; site deerflow.tech
  • Verifiable subagent delegation + tamper-evident tool stamps; optional durable batch_task
  • Pluggable memory.manager_class; FTS5/BM25; OpenViking & mem0 HTTP backends
  • Sandbox: E2B replicas cap; BoxLite / Tenki / OpenSandbox among new providers
  • Breakings: always-on X-Trace-Id, PAT auth, loopback Docker bind — read CHANGELOG before upgrade
Read full story
Linked Models & Agents:
⚡ Doubao Seed 2.0 Code⚡ DeepSeek-v3.2⚡ Kimi-2.5🤖 DeerFlow🤖 SuperAgent🌐 ByteDance🌐 Volcengine
#DeerFlow#ByteDance#SuperAgent#Sandbox#Memory#MCP#Subagents#PAT#E2B
L
LangChain@langchain-ai·
👀75
🔥90

LangGraph.js SDK 1.12.0: useStream enqueue can use durable server-side run queues across reloads/sessions

LangGraph.js shipped @langchain/[email protected] on 2026-09-25 (with @langchain/[email protected] and Vue/Svelte twins). useStream's "enqueue" multitask strategy can set queue: "server" so queued runs hit backend Runs REST instead of an in-memory client queue—surviving reloads, visible across sessions, with server-side cancelQueued/clearQueue. Default remains queue: "local"; custom AgentServerAdapter transports are unsupported. For frontends that need durable, collaborative agent queues.

LangGraph.js SDK 1.12.0: useStream enqueue can use durable server-side run queues across reloads/sessions

Key Takeaways

  • Shipped: langgraph-sdk 1.12.0 + react/vue/svelte 1.2.0
  • Opt in with queue: "server" on useStream enqueue; local remains default
  • Backend must implement Runs REST create/list/cancel endpoints
  • Docs: useStream React howto + npm package
  • Not supported with custom AgentServerAdapter transports
Read full story
Linked Models & Agents:
🤖 LangGraph-Agent🤖 useStream🌐 LangChain🌐 LangGraph
#LangGraph#LangGraphJS#useStream#React#Durable queue#Agent-UI#TypeScript
Q
QwenLM@Alibaba_Qwen·
👀80
🔥93

Qwen Code 0.24.6: Managed Runtime v2, native Advisor tool, /batch-api, Hosted & Durable Sessions

Alibaba’s open-source terminal coding agent Qwen Code shipped v0.24.6 on 2026-09-26 (npm @qwen-code/qwen-code 0.24.6; Desktop desktop-v0.24.6 and SDK TypeScript sdk-typescript-v0.1.16 same day): CLI declares Managed Runtime v2 execute/status/cancel and mounts tool ops on the worker; adds a native Advisor consultation tool with usage limits; /batch-api for agent-prepared batch workflows; serve gains guarded Hosted Runtime, Spring control plane + dual-path WebShell, and model/reasoning selection at session creation; core adds Durable Managed Session journal and failover. Release notes list no known breaking changes.

Qwen Code 0.24.6: Managed Runtime v2, native Advisor tool, /batch-api, Hosted & Durable Sessions

Key Takeaways

  • Shipped: v0.24.6 / npm 0.24.6; Desktop desktop-v0.24.6 same window
  • Runtime: Managed Runtime v2 execute/status/cancel contract; tool ops mounted on the worker
  • Advisor: native advisor tool with consultation behavior and usage limits
  • Batch: /batch-api agent-prepared Batch API workflow
  • Hosted: guarded Hosted Runtime + Durable Managed Session journal/failover; model/reasoning at session create; no known breaking changes
Read full story
Linked Models & Agents:
⚡ Qwen🤖 Qwen Code🌐 alibaba🌐 Qwen (通义千问)
#Qwen Code#Managed Runtime#Advisor#Batch API#Hosted Runtime#coding agent#Alibaba
ADSponsored
P
Pydantic@pydantic·
👀73
🔥92

Pydantic AI 2.50.0: DecisionModel protocol, named route selection, Gemini 3.8 Live realtime

Pydantic AI 2.50.0 (2026-09-25) adds DecisionModel (TypeSafeModel implements it), named route questions with optional decision_route_threshold, Gemini 3.8 Live realtime, durable-context hooks, and audio_seconds usage pricing.

Pydantic AI 2.50.0: DecisionModel protocol, named route selection, Gemini 3.8 Live realtime

Key Takeaways

  • DecisionModel base + TypeSafeModel implementer; ask routes by name (#8696, #8733)
  • Opt-in decision_route_threshold replaces tool-call lean; ModelSelectionContext.prompt (#8737/#8739)
  • Gemini 3.8 Live realtime + RequestUsage.audio_seconds (#8393/#8427)
  • decide spans + RunContext.in_durable_context (#8698/#8723)
  • Upgrade: pip install -U 'pydantic-ai==2.50.0'
Read full story
Linked Models & Agents:
⚡ Gemini 3.8 Live⚡ Jev🤖 Pydantic AI🌐 Pydantic🌐 Google
#Pydantic AI#DecisionModel#TypeSafe#Gemini Live#Durable#Python
M
Mastra@mastra_ai·
👀55
🔥90

Mastra @mastra/core 1.70.0: ModelSelectionProcessor routes cheap vs strong models; durable streams gain closeOnSuspend

Mastra published @mastra/[email protected] on 2026-09-24 with ModelSelectionProcessor for classifier-backed per-request model routing (#24865), closeOnSuspend on durable agent streams so AG-UI-style loops can finish on tool suspend, plus createClassifierScorer(), optional tool titles, and trace aggregate dimension/measure helpers with tag predicates. Docs warn routing is not always cheaper because models do not share prompt caches.

Key Takeaways

  • ModelSelectionProcessor routes simple vs hard requests (#24865); scope: 'first-step' optional
  • Durable closeOnSuspend: true ends streams at tool suspend (#24894)
  • createClassifierScorer() maps Classifier questions to 0–1 scores (#24792)
  • Trace aggregate descriptors + tag predicates for advanced queries
  • Upgrade: npm i @mastra/[email protected]; see CHANGELOG + processors docs
Read full story
Linked Models & Agents:
🤖 Mastra🌐 Mastra
#Mastra#TypeScript#ModelSelectionProcessor#durable agents#Classifiers#Observability
D
Dan Kornas@DanKornas·
👀68
🔥91
View on 𝕏 ↗

Open Managed Agents open-sources a self-hosted managed-agent platform for Node, Docker, or Cloudflare

@DanKornas highlighted Open Managed Agents (github.com/openma-ai/open-managed-agents), an Apache-2.0 self-hosted platform for builders who want managed-agent workflows without vendor lock-in: durable sessions, sandboxed tools, memory, encrypted credentials, and crash recovery on Node+Docker or Cloudflare Workers, with Claude Managed Agents API and OpenAI SDK paths via /openai/v1.

Key Takeaways

  • Self-host managed-agent workflows with durable sessions, memory, and crash recovery.
  • Supports sandboxed tools, private tools, and MCP; bring your own model keys.
  • Runs on Node/Docker or Cloudflare Workers; Claude Managed Agents API and OpenAI SDK compatible.
Read full story
Linked Models & Agents:
🤖 Claude Code (Anthropic)🌐 Anthropic🌐 OpenAI🌐 cloudflare
#Open Managed Agents#Self hosted#MCP#Cloudflare Workers#Agents#open source
D
Dan Kornas@DanKornas·
👀50
🔥92
View on 𝕏 ↗

agmsg open-sources local SQLite messaging for CLI coding agents

@DanKornas introduced agmsg, a local messaging tool for Claude Code, Codex, Gemini CLI, and GitHub Copilot CLI. Independent agent sessions coordinate through one shared SQLite store without a daemon, network service, or human copy-paste relay; durable history can be replayed into a fresh session, with monitor, turn, both, and manual delivery modes. The Bash + sqlite3 project is MIT-licensed and offers npx, plugin, and direct-script installation paths.

Key Takeaways

  • A shared local SQLite transport lets independent CLI agent sessions hand work to one another directly.
  • Durable history can be replayed into fresh sessions, with monitor, turn, both, and manual delivery modes.
  • Bash + sqlite3, MIT open source, covering Claude Code, Codex, Gemini CLI, and Copilot CLI.
Read full story
Linked Models & Agents:
🤖 Claude Code (Anthropic)
#Agmsg#CLI agents#SQLite#Claude Code#Codex#Gemini CLI#Copilot CLI#MIT#open source
T
Temporal@temporalio·
👀38
🔥96
View on 𝕏 ↗

Temporal Raises $550M Series E at a $12.55B Valuation for Durable Agent Workflows

Temporal announced a $550M Series E at a $12.55B valuation. Its durable-execution infrastructure is designed to let AI agents resume multi-step workflows after crashes, bugs, or long waits, targeting reliability and orchestration for production agents.

Key Takeaways

  • The Series E totals $550M at a $12.55B valuation
  • Durable execution preserves long-running workflow state so agents can recover from failures or waits
  • Production agents increasingly depend on reliable execution, retries, and orchestration beyond raw model capability
Read full story
Linked Models & Agents:
🤖 agent-runtime🌐 temporal
#Temporal#AI Agents#Durable Execution#Workflow Orchestration#Series E