--- id: 2026-w45 schema_version: 1 issue_number: 2 title: "Frontier labs converge on a shared agent-interop spec" date_start: 2026-10-31 date_end: 2026-11-06 pages: - cover - briefing - models - performance - launches - funding - deals - markets - discourse - most_covered - back_page sources_total: 34 source_counts: "Exa Markets": 8 "Launch Index": 4 "Third Run Time": 2 "The Morning Build": 2 "Microsoft AI": 2 "VentureBeat": 2 "AI Atlas": 2 "SiliconReport": 2 "Funding Index": 2 "AI Lately": 1 "Creati AI": 1 "Business Wire": 1 "PR Newswire": 1 "GlobeNewswire": 1 "CBC Globe": 1 "Reuters": 1 "Associated Press": 1 markdown: /2026-w45.md json: /2026-w45.json --- # Frontier labs converge on a shared agent-interop spec *The Weekly AI Edition · № 2 · Oct 31–Nov 6 (2026-10-31 → 2026-11-06)* Anthropic, OpenAI and Google DeepMind publish a joint draft for cross-vendor agent identity, tool permissions and provenance headers — the week agents started speaking one wire protocol. ## The Briefing The week belonged to interop: three frontier labs published a joint draft for agent identity and tool-permission headers, and every major agent framework committed to implement it. OpenAI released GPT-6.1 Terra, a mid-tier computer-use model priced at half its predecessor, and said its models produced three mathematical results including a claimed refutation of the Exponential Time Hypothesis. Within hours, Anthropic claimed a refutation of its own, leaving outside shepherds to decide whether either proof survives review. Google DeepMind previewed Genie 4, a world model that generates interactive scenes with sound, while Meta open-sourced the 816-billion-parameter Muse Spark Open for agentic coding. The money kept moving: Tenstorrent raised a $1.5 billion Series E at a $6.2 billion valuation, and Guidewire agreed to acquire hyperexponential for $1.7 billion in cash. Underneath the rally, investors kept debating whether debt-funded AI capital expenditure can pay for itself. ## Models & Releases ### GPT-6.1 Terra OpenAI's mid-tier computer-use model lands at $1 per million input tokens — half the price of GPT-5.6 Terra and more than three times faster. > The price cut, not the benchmark, is the story. Companies: `openai` Sources: [AI Lately](https://ailately.com/rn/2026-10-27-12) · [Third Run Time](https://www.thirdruntime.com/?date=2026-10-26) ### Genie 4 Google DeepMind's world model generates interactive scenes with native audio and real-time steering, in research preview for trusted testers. Companies: `google` Sources: [The Morning Build](https://themorningbuild.com/digest/2026-10-28/) ### Muse Spark Open Meta's 816-billion-parameter open-weight MoE targets long-running coding agents, with 42 billion parameters active per token and a one-million-token context. Companies: `meta` Sources: [Creati AI](https://creati.ai/ai-news/2026-10-27/meta-releases-muse-spark-open-an-816b-parameter-open-weight-model-for-multimodal-coding-agents/) ### MAI Image 2.7 Microsoft AI's latest image editing model debuts at No. 3 on Arena Edit and No. 2 on Artificial Analysis's image editing leaderboard. Companies: `microsoft` Sources: [Microsoft AI](https://microsoft.ai/news/mai-image-2-7-debut/) ### Liquid AI LFM3 Six compact models — language, vision, speech, structured decisions — built to run enterprise agents on company-controlled hardware. Companies: `liquid-ai` Sources: [VentureBeat](https://venturebeat.com/technology/liquid-ais-new-lfm3-models-put-browser-agents-voice-commands-and-image-recognition-onto-enterprise-controlled-devices) ## Performance ### GPT-6.1 Terra on Terminal-Bench 4.0 **`50.3%`** Company-reported score more than doubles its predecessor; no independent evaluation had been published by Monday night. Evidence: "scores 50.3% on Terminal-Bench 4.0 at max effort" — AI Atlas Companies: `openai` Sources: [AI Atlas](https://atlas.prashish.com/weeks/2026-10-26) ### MAI Image 2.7 on Arena Edit **`#3`** Microsoft's image editor debuts third on Arena Edit, making Microsoft AI the No. 2 lab on the board. Evidence: "launches in preview at No. 3 on Arena Edit" — Microsoft AI Companies: `microsoft` Sources: [Microsoft AI](https://microsoft.ai/news/mai-image-2-7-debut/) ### LFM3-1.7B on WebArena-Verified **`38.42`** Liquid's smallest text model claims browser-agent results above substantially larger rivals on selected tool-use tasks. Evidence: "LFM3-1.7B scores 38.42 on WebArena-Verified" — VentureBeat Companies: `liquid-ai` Sources: [VentureBeat](https://venturebeat.com/technology/liquid-ais-new-lfm3-models-put-browser-agents-voice-commands-and-image-recognition-onto-enterprise-controlled-devices) ## Launches ### Claude Duet A proactive Anthropic agent that understands what needs your attention — 322 votes on Product Hunt this week. Companies: `anthropic` Sources: [Launch Index](https://launchindex.ai/month/2026-10) ### 1Password for Agents Lets AI agents sign in for you without ever seeing your passwords. Companies: `1password` Sources: [Launch Index](https://launchindex.ai/day/2026-10-24) ### Mistral Voxtral 3 Mistral's speech models for transcription, text-to-speech, and agents ship on Product Hunt. Companies: `mistral` Sources: [Launch Index](https://launchindex.ai/day/2026-10-22) ### ElevenLabs Realtime Translation Turns live speech into translated audio as people talk. Companies: `elevenlabs` Sources: [Launch Index](https://launchindex.ai/day/2026-10-22) ## Funding ### Anthropic files to go public **`$131.7B`** · IPO filing The filing discloses $131.7 billion in cash and investments; the paperwork alone was enough to lift AI stocks for a third straight session. Evidence: "Anthropic discloses $131.7 billion in cash and investments" — SiliconReport Companies: `anthropic` Sources: [SiliconReport](https://www.siliconreport.com/capital/archive/2026-10) · [The Morning Build](https://themorningbuild.com/digest/2026-10-28/) ### Tenstorrent raised $1.5B Series E **`$1.5B`** · Series E · led by Innovation X Jim Keller's RISC-V AI compute company reaches a $6.2 billion valuation as it scales Galaxy server production and IP licensing. Evidence: "co-led Tenstorrent's $1.5 billion Series E financing at a $6.2 billion post-money valuation" — Business Wire Companies: `tenstorrent` Sources: [Business Wire](https://www.businesswire.com/news/home/20261027228454/en/Innovation-X-Co-Leads-Tenstorrents-%241.5-Billion-Series-E-at-a-%246.2-Billion-Valuation) ### General Compute raised $100M Series A **`$100M`** · Series A Specialized cloud infrastructure and chips for AI inference, at an $800 million valuation. Evidence: "General Compute raises $100 million Series A" — SiliconReport Companies: `general-compute` Sources: [SiliconReport](https://www.siliconreport.com/capital/archive/2026-10) ### Hyphenate raised $45M Series A **`$45M`** · Series A · led by Nexus Venture Partners Autonomous AI agent systems for enterprise financial operations. Evidence: "Hyphenate raises $45M Series A" — Funding Index Companies: `hyphenate` Sources: [Funding Index](https://funding.fru.dev/months/2026-10) ### Infisical raised $40M Series B **`$40M`** · Series B · led by Elad Gil Secrets-management infrastructure for enterprise systems and AI agents, at a $600 million valuation. Evidence: "Infisical raises $40M Series B" — Funding Index Companies: `infisical` Sources: [Funding Index](https://funding.fru.dev/months/2026-10) ## Deals ### Guidewire → hyperexponential **`$1.7B`** The P&C insurer platform buys the London-based agentic underwriting specialist, setting up an 'agentic underwriting' push across commercial lines. Evidence: "base purchase price of $1.7 billion in cash" — PR Newswire Companies: `guidewire` `hyperexponential` Sources: [PR Newswire](https://www.prnewswire.com/news-releases/guidewire-signs-definitive-agreement-to-acquire-hyperexponential-302920513.html) ### RapidCanvas → Entytle **`undisclosed`** The managed AI-execution company absorbs the Autonomous Aftermarket pioneer and its Insyghts platform, a week after closing a $29M Series B. Companies: `rapidcanvas` `entytle` Sources: [GlobeNewswire](https://www.globenewswire.com/news-release/2026/10/28/3390266/0/en/rapidcanvas-acquires-entytle-to-bring-ai-driven-execution-to-the-industrial-aftermarket.html) ## Markets **AI basket index, week to date** (index) | Date | Value | | --- | ---: | | 2026-10-23 | 100 | | 2026-10-26 | 101.5 | | 2026-10-27 | 103 | | Ticker | Name | Close | WoW | | --- | --- | ---: | ---: | | NVDA | NVIDIA | 217.94 | +3.31% | | MSFT | Microsoft | 531.33 | +1.3% | | GOOGL | Alphabet | 320.58 | +0.12% | | META | Meta Platforms | 620.19 | -3.49% | | AMD | AMD | 560.26 | +8.41% | | TSM | TSMC | 436.87 | +4.39% | | AMZN | Amazon | 253.1 | +2.06% | | ARM | Arm Holdings | 226.6 | +7.73% | - NVDA: "2026-10-27 USD 217.94" — Exa Markets - MSFT: "2026-10-27 USD 531.33" — Exa Markets - GOOGL: "2026-10-27 USD 320.58" — Exa Markets - META: "2026-10-27 USD 620.19" — Exa Markets - AMD: "AMD closed at $560.26, up $28.98 (5.45%)" — CBC Globe - TSM: "2026-10-27 USD 436.87" — Exa Markets - AMZN: "2026-10-27 USD 253.10" — Exa Markets - ARM: "2026-10-27 USD 226.60" — Exa Markets Data: Exa Markets · CBC Globe Sources: [Exa Markets](https://exa.ai/library/markets/stock/NVDA?date=2026-10-27&view=series&from=2026-10-19) · [Exa Markets](https://exa.ai/library/markets/stock/MSFT?date=2026-10-27&view=series&from=2026-10-19) · [Exa Markets](https://exa.ai/library/markets/stock/GOOGL?date=2026-10-27&view=series&from=2026-10-19) · [Exa Markets](https://exa.ai/library/markets/stock/META?date=2026-10-27&view=series&from=2026-10-19) · [CBC Globe](https://cbcglobe.com/en/articles/2026/10/27/breaking-tesla-surges-4-and-amd-over-5-on-nyse-nvidia-and-mi-4ca4b69e) · [Exa Markets](https://exa.ai/library/markets/stock/AMD?date=2026-10-27&view=series&from=2026-10-19) · [Exa Markets](https://exa.ai/library/markets/stock/TSM?date=2026-10-27&view=series&from=2026-10-19) · [Exa Markets](https://exa.ai/library/markets/stock/AMZN?date=2026-10-27&view=series&from=2026-10-19) · [Exa Markets](https://exa.ai/library/markets/stock/ARM?date=2026-10-27&view=series&from=2026-10-19) ## Discourse ### Did two frontier labs just refute the Exponential Time Hypothesis — or can nobody actually check the proofs? OpenAI says its models found sub-exponential algorithms for a hard class of logic puzzles; Anthropic reportedly claimed its own refutation hours later. Outside shepherds are organising expert groups to verify each result. Companies: `openai` `anthropic` Sources: [AI Atlas](https://atlas.prashish.com/weeks/2026-10-26) · [Third Run Time](https://www.thirdruntime.com/?date=2026-10-26) ### Can debt-funded AI capital expenditure pay for itself? Relief rally or reckoning: Anthropic's IPO paperwork lifted chip stocks all week, while investors keep asking whether borrowing-fuelled data-centre spend earns its cost of capital. Companies: `anthropic` `meta` Sources: [Reuters](https://finance.yahoo.com/markets/stocks/articles/us-stock-futures-muted-oil-095309821.html) · [Associated Press](https://apnews.com/article/stocks-markets-ai-earnings-oil-e11a2240fc3283c4507d44b9833f30a2) ## Most Covered 1. **OpenAI** (`openai`) — 3 stories, trending up 2. **Anthropic** (`anthropic`) — 3 stories, trending up 3. **Liquid AI** (`liquid-ai`) — 2 stories, trending up 4. **Meta** (`meta`) — 2 stories, trending down 5. **Google** (`google`) — 1 story, trending flat ## Back Page **Sources** | Outlet | Items | | --- | ---: | | Exa Markets | 8 | | Launch Index | 4 | | Third Run Time | 2 | | The Morning Build | 2 | | Microsoft AI | 2 | | VentureBeat | 2 | | AI Atlas | 2 | | SiliconReport | 2 | | Funding Index | 2 | | AI Lately | 1 | | Creati AI | 1 | | Business Wire | 1 | | PR Newswire | 1 | | GlobeNewswire | 1 | | CBC Globe | 1 | | Reuters | 1 | | Associated Press | 1 | 34 fetched source records this week. **How this issue was made** Ingested via deterministic RSS/API adapters, normalized, deduplicated, classified, ranked, extracted, and drafted by the Weekly AI Edition pipeline, then verified. Every item carries at least one fetched source record; every number carries an evidence quote from the stored excerpt. **For agents** - Markdown: /2026-w45.md - JSON: /2026-w45.json - llms.txt: /llms.txt - Feed: /feed.xml Code MIT · Content CC BY-NC with sources attributed.