Search “Meta”

Sierra announces ARR surpassing $400 million

On October 30, 2026, Sierra announced on its official blog that the company's annualized revenue (ARR) had surpassed $400 million. Sierra said that in the third quarter alone it added more than $150 million in ARR, exceeding its ARR for the entire previous year. The company also said it took six months for ARR to grow from $100 million to $200 million, and only five months to grow from $200 million to $400 million, and that it currently serves nearly 50% of Fortune 50 companies. In the same blog post, Sierra mentioned that a large U.S. retailer with revenue exceeding $40 billion used Ghostwriter to go from first meeting to agent launch in 27 days, and said it is working with more than 50 companies including Meta, Shopify, Stripe, and Walmart to advance Poppy, an open standard for personal agent interaction.

Rohan Paul says Anthropic's revenue curve will become legendary

On October 28, 2026, Rohan Paul offered the judgment that Anthropic's revenue curve will become legendary. Citing a Bloomberg chart, he said that Anthropic's revenue in the first half of 2026 was nearly 400 times its full-year 2023 revenue, and that no other company in the chart comes close to that multiple; Nvidia, the second-fastest grower, remained below 50 times throughout its first three years after going public, while Alphabet, Meta, Microsoft and SpaceX essentially stayed near the 1x baseline. The same material also states that Anthropic's full-year 2025 revenue was about 100 times that of 2023, so its revenue in just six months of 2026 was about 4 times its full-year 2025 revenue. At present, this judgment appears only in this single source, Rohan Paul, and there have been no subsequent developments.

ARC Prize announces the evaluation results of Muse Spark Open on ARC-AGI

On October 29, 2026, ARC Prize announced the verified results of Meta's Muse Spark Open on ARC-AGI: 89.0% on ARC-AGI-1 at $0.29 per task, 57.5% on ARC-AGI-2 at $0.73 per task, and 0.1% on ARC-AGI-3 under the standard harness at $3.9K per task. ARC Prize stated that its v1/v2 scores are close to those of Haiku 5.5 and GPT-6 Luna, but at roughly 9 times the cost of the latter; the model is priced on OpenRouter at $1.25 per million input tokens and $4.25 per million output tokens. As the reasoning tier increases, ARC-AGI-2 rises from 29.4% at medium (an average of 55k tokens per task) to 57.2% at xhigh (180k tokens per task).

Nathan Lambert criticizes models for being overly evasive on gray-area questions

On October 30, 2026, Nathan Lambert commented on Simon Taylor's post about the answering style of Meta Muse, calling this style "very necessary." He said that discussing intersectional issues in fields such as law and medicine with models like Claude is a headache: the models are eager to hedge risks and avoid gray areas, which often ends up sidestepping the real question. There have been no further developments so far.

Artificial Analysis Announces AA-PersonalityBench Friendliness Rankings

On October 29, 2026, Artificial Analysis announced the Friendliness rankings for the AA-PersonalityBench evaluation. In this evaluation, Meta AI's Muse series models received the highest friendliness scores: Muse Spark 1.3 scored 100, Muse Spark Open 98, Muse Wonder 95, and Muse Glimmer 93. Friendliness measures whether responses convey warmth and approachability without seeming artificial. The lowest friendliness scores were Nemotron 3 Ultra (1 point), GPT-3.5 Turbo (3 points), MiniMax-M3.1-Flash (6 points), Grok 4.6 (8 points), and DeepSeek V4.1 Flash 1017 (11 points). There have been no subsequent developments so far.

Hebrew University team publishes BBERT bacterial DNA identification research

Researchers from the Hebrew University and other institutions have published a paper titled "Fast and accurate taxonomic domain assignment of short metagenomic reads using BBERT," proposing the BBERT model. This model can identify bacterial sequences in DNA fragments as short as 100 bases; in benchmark tests, its accuracy in distinguishing bacterial from non-bacterial DNA is about 90%, its accuracy in identifying protein-coding fragments is about 88%, and its accuracy in determining the reading frame is about 96%. In simulated mixed samples, filtering with BBERT first reduced the average assembly time from about 66 hours to 25 hours while retaining about 96% average bacterial genome coverage. The research has currently been published as a paper, and no follow-up progress has been reported.

Researchers from Jagiellonian University and other institutions have published the DomeGS paper

English text is being prepared.

Infini-AI-Lab releases the open-source harness adapter framework Hermit

On October 25, 2026, Infini-AI-Lab released the open-source framework Hermit, which adapts AI agents by training external harness adapters while keeping the model and harness frozen. According to a MarkTechPost report, version v0.1.0 of the framework is implemented in Python under the Apache-2.0 license and can be installed with pip install hermit-harness; it includes five built-in evolution methods: GEPA, SkillOpt, MetaWorld Model, Harness2, and Continual Harness. The report states that on GLM 5.3 Flash, the MWM adapter achieved 0.795 accuracy on the CL-Bench database exploration task, higher than the in-context baseline of 0.580, but most results are self-reported data from single runs. There are currently no subsequent developments.

Hebrew University team publishes BBERT bacterial DNA identification research

Researchers from the Hebrew University and other institutions have published a paper titled "Fast and accurate taxonomic domain assignment of short metagenomic reads using BBERT," proposing the BBERT model. This model can identify bacterial sequences in DNA fragments as short as 100 bases; in benchmark tests, its accuracy in distinguishing bacterial from non-bacterial DNA is about 90%, its accuracy in identifying protein-coding fragments is about 88%, and its accuracy in determining the reading frame is about 96%. In simulated mixed samples, filtering with BBERT first reduced the average assembly time from about 66 hours to 25 hours while retaining about 96% average bacterial genome coverage. The research has currently been published as a paper, and no follow-up progress has been reported.

View story:Hebrew University team publishes BBERT bacterial DNA identification research

Goldman Sachs' Hatzius says growth in AI data center investment will peak

Goldman Sachs Chief Economist Jan Hatzius said that the growth in AI investment by hyperscale data centers is expected to peak in the coming weeks, and its year-on-year change as a share of U.S. GDP is expected to peak in the fourth quarter of 2026 at about 1.5%. Related reports mentioned that Amazon, Microsoft, Alphabet, Meta, Oracle and SpaceX are expected to spend about $725 billion to $800 billion on AI infrastructure capital expenditure this year, an increase of 64% over 2025. Goldman Sachs also provided scenario forecasts: under the optimistic scenario, AI-related stocks rise 55% and the broad market rises 30%; under the pessimistic scenario, the broad market falls 20%; if the data center investment boom ends, the level of U.S. GDP could decline by about 2 percentage points. There are currently no subsequent developments.

Apollo Chief Economist Torsten Slok: No AI Financial Returns Seen Yet Outside US Tech Industry

On October 24, 2026, Apollo Chief Economist Torsten Slok put forward a judgment in the "Daily Spark" column: AI has not yet brought financial returns outside the US tech industry. He pointed out that, after adjusting for inflation, revenue per employee at the seven tech giants—Nvidia, Tesla, Amazon, Microsoft, Alphabet, Apple, and Meta—is rising, while both revenue per employee and profit margins at the other 493 companies in the S&P 500 have remained flat, and there is no sign of AI driving productivity growth. Slok also said that mainstream enterprises have begun to cut back on AI spending, treating token consumption as a cost that drags down profit margins rather than a source of growth; he believes the ultimate direction of the AI narrative will be decided by the S&P 493. Subsequent reports repeated the above views and mentioned that Goldman Sachs economist Jan Hatzius holds the opposite view. There are currently no new developments.

The author judges that Apple Glasses will only be sold to people who already wear glasses

On October 24, 2026, 9to5Mac author Ben Lovejoy offered a judgment, arguing that Apple Glasses will only sell to people who already have a habit of wearing glasses. Based on his experience with the Ray-Ban Meta and Even Realities G2 smart glasses, he pointed out that Apple Glasses have limited appeal for people who do not wear glasses. According to his description, the first generation is expected to be released next year, with no display and only audio output, and the camera used only for Visual Intelligence; the second generation will have a display, with navigation and translation being the most practical scenarios, but the frequency of demand is not high. On this basis, he judges that people who wear glasses and have a sufficient budget may want the second generation, while those who do not wear glasses will be hard to persuade. At present, this judgment remains the author's personal opinion, and there are no subsequent factual developments.

Carl's AI Watts shares Personal Agent entry rankings

On October 23, 2026, Carl's AI Watts published an article ranking 13 Personal Agents by the number of apps they can connect to. According to the article, meta muse tops the list with 54, chatgpt dots has 50, gemini spark has 35, today and poke each have 30, grok bot has 20, manus cue has 15, Instinct has 12, and hark, Lingguang 2.0, Qwen Go, Hermes, and OpenClaw each have about 10. The article also notes that Hark is free with no feature limits, muse is only available in the US and Canada, and Qwen Go, Lingguang 2.0, cue, and Instinct still require invitation-only beta access. This ranking is currently only seen in this source, with no subsequent developments yet.

University of Chicago paper points out that LLMs' self-reported key decision factors are unreliable

On October 24, 2026, the University of Chicago published a paper pointing out that large language models (LLMs) cannot reliably self-report the key factors in their own decisions. The paper states that LLMs cannot predict the key questions on which their own answers diverge, and that the key decision factors they self-report match the true factors only 13.4% of the time, lower than the 39.2% achieved after random shuffling. The study had models repeatedly predict 59 March Madness games and 15 Metaculus events, then used a reranker to compare answer pairs, with computation about 86 times lower than local LLM judging. The proportion of individual answers covering the true key factors was 28.6%, versus a baseline of 37.8%. Currently this finding has been reported by a single source, with no subsequent developments yet.

Artificial Analysis publishes AA-AgentPerf results for Muse Spark Open

Artificial Analysis on October 24, 2026 published on AA-AgentPerf preliminary test results for Meta's newly open-weight model Muse Spark Open. The model is a 816B-parameter, 42B-active MoE, scoring 39 on the Artificial Analysis Intelligence Index. Tests show that NVIDIA B300 and GB300 each reach about 150M tokens/s per megawatt of power under throughput-first configurations; GB300 reaches about 108M and B300 about 92M at 100 tok/s output speed; AMD MI355X reaches 70M tokens/s per megawatt at 65 tok/s. Of the model's 62 layers, 46 use 2,048-token sliding window attention, reducing KV cache requirements at 1M context by about 75% versus full attention. Meta released an NVFP4 checkpoint, and AMD released an MXFP4 checkpoint. There are currently no subsequent development facts.

Nomura strategist says market is rapidly repricing left-tail risk

Nomura cross-asset strategist Charlie McElligott proposed the judgment that the market is rapidly repricing left-tail risk, with the driving force shifting from the inflation narrative to a systemic revaluation of AI-driven growth prospects, and with long-duration Treasury positions once again serving as a hedge for equity portfolios. His report noted that overnight the 10-year Treasury yield plunged more than 15 basis points in a single day, at one point falling below 5.00%, the Nasdaq fell 2.60%, and the Philadelphia Semiconductor Index fell 6.72%; he attributed the root cause of the selloff to a rethink of the AI trajectory, and mentioned that Anthropic's annualized recurring revenue data disappointed the market, that Gartner warned more than 80% of emerging cloud service providers face bankruptcy, and that Meta's CDS spread widened from 80 basis points on July 29 to 120 basis points on October 22. An October 24 report further stated that McElligott pointed out that hyperscale cloud providers' AI capital expenditure is caught in a dilemma: continuing to accelerate would put pressure on free cash flow, while a substantial slowdown would mean that AI demand and earnings expectations have failed; in the most violent single-day selloff of the past week, 47 of the 124 US equity long-short pairs monitored by Nomura moved more than 3 standard deviations, and 6 moved more than 5 standard deviations, while over the past two weeks CTA trend strategies have closed out about $68 billion in Treasury futures shorts and $38 billion in short-term interest rate futures shorts.

OpenClaw says cybersecurity is developing faster than AI this week

On October 23, 2026, OpenClaw reposted and commented on the "personal AI agents" weekly roundup compiled by McBain, and offered this judgment: the only industry changing faster than AI this week is cybersecurity. The weekly roundup covers Anthropic, OpenAI, Meta, OpenClaw, Perplexity, and others. OpenClaw's comment made the above comparison in a teasing tone; there is currently no further follow-up development or independent source cross-verification.

Warp Sciences raises $215 million to build AI chemistry research lab

Warp Sciences has completed a $215 million financing round to build a chemistry research facility in the United States, using AI and robotics to assist pharmaceutical and biotechnology companies in developing drugs. The company was founded by Kevin Weil, former head of OpenAI's science initiatives, about six months after he left OpenAI; its co-founders include former White House chief data scientist DJ Patil and former Meta executive Mano Paluri. The company operates as a chemistry contract research organization (CRO), initially focusing on lead compound optimization through analog synthesis. Experimental data will be used to train its AI models, while chemists wear cameras and sensor-equipped gloves to demonstrate actions for robots to learn experimental procedures. The round was led by Lightspeed Venture Partners, with participation from Premji Invest, Mayfield, IVP, Lux, Conviction, Menlo Ventures, 01 Advisors and others; angel investors include Brian Armstrong, Ion Stoica and Dylan Field. No further developments have been disclosed at this time.

1Password launches password login feature for AI agents

On October 22, 2026, 1Password launched 1Password for Agents, allowing AI agents to complete logins without viewing passwords. According to 9to5Google, when an agent reaches a login form, it authenticates to 1Password using a digital key; after the user confirms the required credentials and their purpose, 1Password fills them directly into the form. The credentials are not handed to the agent and are removed once login succeeds, so the agent cannot read or reuse them. The integration currently supports only Meta's Muse and Grok Bot, with OpenAI Dots and Anthropic Claude Duet to follow. Another report says the first integrations are Meta's Muse and SpaceXSI's Grok Bot, to be deployed within the next few days; partners including Anthropic's Claude Duet, OpenAI's ChatGPT Work and dots, and Perplexity have not yet given a timeline. The service is included in personal and family subscriptions at no extra charge and currently supports only accounts hosted on 1Password.com; 1Password.eu and 1Password.ca accounts are not yet supported.

Google open-sources TorchTPU native PyTorch stack

Google collaborated with Meta's PyTorch team to launch the open-source project TorchTPU, providing a native PyTorch backend for Google TPU; migration only requires changing model.to("cuda") to model.to("tpu"). The project supports both eager and torch.compile modes, adds a new DeferAndFuse eager mode, covers about 450 Aten operators, and has been verified to run out of the box on 4,917 Hugging Face models, with a Tier 1 pass rate of 98% and Tier 2 numerical consistency relative to GPU of 97%. The official statement says it has been used in the inference and training production workloads of some Google Cloud major customers, and is integrated with vLLM, SGLang, TorchTitan, Ray, and TorchRec. Subsequent reports said that at PyTorch Conference North America 2026, Google open-sourced TorchTPU, whose runtime no longer depends on JAX, and existing workloads can be migrated with only minor code changes; PyTorch noted that TorchTPU's default Skinny Collectives scheme suffers roughly 30–40% end-to-end performance loss under distributed workloads compared with optimized schemes, and achieving peak performance still requires Pallas kernels and vLLM-T