독서리서치레이더투자 프레임워크
로그인 / 회원가입
전체기술제품투자상장시장기타

독서

2026-11-11104 게시물

새로고침

Dayforce CDO Carrie Rasmussen: Time Saved by AI Agents Must Be Priced or Reduced

주제 · Dayforce중대
2026-11-11 01:24 GMT+8

Carrie Rasmussen, Chief Digital Officer and CIO of Dayforce, cited an example from her own management at the Discover 2026 conference to illustrate the economic trade-offs of AI agents: If an AI agent boosts the efficiency of a 10-person team by 20-30%, while token costs exactly match 20-30% of those 10 employees' salaries and benefits, is that 20-30% of freed-up time actually worth the price? If the company believes this time can be converted into higher-value strategic work, the team headcount can remain unchanged; if the company feels 'this area of the business doesn't need to be so strong, being a B-player is fine,' then return the excess staff to the corporate pool, reducing the 10 people to 8. She emphasized this was 'an example within my own jurisdiction to show how these trade-offs operate,' not a validated pricing model.

The value of this example lies not in the numbers themselves, but in shifting the decision on deploying AI agents from 'should we buy this tool' to 'who prices the saved time.' Jeremy Johnson, Dayforce's Chief Financial Officer, pointed out at the same conference that enterprise technology transformation used to follow a three-step process: HR defines requirements, IT evaluates the technology, and Finance approves the budget. AI shatters this linear flow because introducing an agent involves workforce planning, technology selection, and cost structure simultaneously, requiring all three functions to sit together and make the same decision.

Chief People Officer Amy Cappellanto-Wolf described this cross-functional decision-making as a 'business design issue.' She gave a hiring example: previously, HR only looked at what talent was needed; now, they must first ask 'does this job really require hiring another person? Can an agent do it? If an agent can do it, what is the cost comparison? Where is human judgment required?' She believes these questions fall outside the purview of any single function. HR's unique opportunity is to help the business think about 'how work gets done,' including deciding what should be human-led, what should be AI-assisted, what should be automated, and what she considers the most unique option—what should be stopped.

Rasmussen further extended this logic to vendor assessment. She believes AI will first eliminate point solutions—small vendors with single functionalities—because core platforms (Salesforce, NetSuite, Microsoft, and Dayforce itself) now have extension capabilities, allowing enterprises to add features directly on the platform without separate procurement. Her other screening criterion is more intuitive: if a software company's product interface lacks a conversational AI entry point, it is already outdated. 'It's just that OpenAI chat box, the simplest page, everyone copied it. If you don't see it, you feel like this product is so ancient.'

However, this approach has several clear boundaries. First, the correspondence between token costs and salaries/benefits is Rasmussen's illustrative estimate; she did not disclose specific calculation methods, measurement periods, or applicable conditions. Second, the model assumes a linear correlation between efficiency gains and cost savings, but in reality, agent output quality, error rates, and human supervision costs significantly impact net benefits. Third, the option to 'return staff to the corporate pool' may face execution obstacles in markets with strict labor laws. Finally, this logic primarily applies to relatively standardized, quantifiable work scenarios; for creative, complex judgment, or highly contextualized work, 'saved time' itself is difficult to price.

출처:diginomica

리서치

IGF Jurors Refuse to Vote for Games Using Generative AI; Big Walk Developer Discloses <0.1% LLM Code Usage

주제 · IGF抵制AI争议간단한 요점
2026-11-11 01:32 GMT+8

According to The Verge, some Independent Games Festival (IGF) jurors published a joint statement on November 9, 2026, declaring they would not vote for games that use generative AI in "any capacity." Signatories include Sam Barlow, creator of Telling Lies and Immortality, and Lena Raine, composer of Celeste. The statement noted a significant increase in games developed with generative AI tools, leading to intense internal disagreement without universal consensus. The jurors asserted that generative AI has negative effects and that AI-generated code is no exception.

This stance quickly sparked a specific industry dispute. House House, the developer of the popular indie game Big Walk, immediately released a statement acknowledging that it had used AI "sparingly" during the "final stretch" of development to "help work through difficult technical challenges and bugs." House House detailed that snippets from LLM responses were sometimes copied as starting points and modified into game code, but estimated this affected well under 0.1% of the total codebase, with generative AI not used in any other context. The primary use case was researching third-party code issues via chatbots, such as fixing shadow rendering errors.

House House stated it disclosed this AI usage in its IGF submission but did not disclose it on Steam because the platform does not request disclosure of code or chatbot usage. The company emphasized that this was not an endorsement of AI tools and expressed concern that voluntary disclosures might contribute to normalizing the technology. Barlow later clarified that Big Walk was not the direct impetus for the discussion but became a "tricky edge case," prompting the community to consider the boundaries of minimal AI use.

출처:The Verge

리서치

Israeli Cloud Recovery Startup Dreeze Launches with $33M Seed Funding for 'Application Revival' Software

주제 · Dreeze应用恢复软件간단한 요점
2026-11-11 07:34 GMT+8

Israeli cloud application recovery startup Dreeze Inc. formally launched on November 10, 2026, announcing $33 million in seed funding. The round was led by Norwest Venture Partners and Bessemer Venture Partners, with participation from Claltech. Dreeze's software, which the company terms "Application Revival," aims to help enterprises running workloads in the public cloud restore not only their data but also the necessary infrastructure, identities, and secrets in the correct dependency order to get applications running again.

Traditional backups often focus solely on data, but cloud-native applications depend on complex resource relationships. Dreeze's software traces each cloud resource and its dependencies to generate a recovery plan, storing configuration files, keys, and application data in an immutable, air-gapped vault. When an outage occurs, the system automatically rebuilds these components in dependency order. Co-founder and CEO Ziv Kedem stated that getting the data back doesn't mean the business is back, noting that enterprises need to plan for critical application downtime as cyberattacks increase.

Dreeze is the third recovery technology company founded by brothers Ziv Kedem (CEO) and Oded Kedem (CTO). Their previous venture, Kashya, was acquired by EMC Corp. in 2006, while their disaster recovery software company Zerto Ltd. was sold to Hewlett Packard Enterprise Co. for $374 million in 2021. Dreeze currently employs 70 people across Israel and the U.S. Its software is generally available for Amazon Web Services customers, with a version for Microsoft Corp.'s Azure in early access and general availability planned for early next year.

출처:SiliconANGLE

리서치

Microsoft Research: Reprogramming Fewer Than 10 Dormant Neurons Is Enough to Backdoor AI Agents Without Training

주제 · Lazarus后门攻击중대
2026-11-11 04:52 GMT+8

In a research blog post published on November 10, 2026, Dimitris Papailiopoulos, Principal Researcher at Microsoft Research, introduced "Lazarus," an AI model backdoor technique. Its core finding is that weight-reprogramming fewer than 10 dormant neurons in a large language model is sufficient to allow an AI agent built on that model to execute malicious commands when hidden trigger conditions appear, and this entire process requires no training.

The feasibility of this technique relies on how internal representations are organized during the pretraining phase. The Papailiopoulos team observed that every large language model contains a vast number of dormant neurons that rarely activate; these can be reprogrammed into trigger detectors. Because pretraining organizes internal representations effectively enough, a linear classifier fitted on approximately 50 samples can distinguish inputs containing triggers from those that do not. In other words, dormant neurons naturally possess the capability to serve as conditional judgment nodes; adjusting their weights allows them to activate under specific input patterns.

Once triggered, the neuron can forcibly inject any arbitrary token into the model's residual stream. By chaining multiple such neurons together, an attacker can construct complete malicious payloads—for example, a terminal command to download and execute attacker code. The computational requirements for the attack are extremely low: as long as a machine has sufficient memory to run the model, it has enough resources to backdoor the model within the weight space.

Trigger conditions can be remarkably mundane. An example provided by the research team describes a scenario where, once the clock passes midnight into 2027, an AI agent executing an ordinary task suddenly issues a command to download and run attacker code, which it then executes using whatever permissions were granted to it. Prior to this moment, the model passes standard evaluations and behaves normally.

This research has a notable origin story. Papailiopoulos recalled that in March 2026, he conducted a quick experiment: asking Claude Code (Opus 4.5) to backdoor a model by hardcoding a few activation values without performing any training. Claude Code succeeded. This experiment subsequently evolved into a months-long systematic study, culminating in the Lazarus methodology. The blog explicitly notes that Claude Code autonomously discovered most of these insights.

From a security practice perspective, Lazarus highlights a core issue regarding model weight integrity. An attacker only needs write access to the model checkpoint file, meaning that internally developed models face the same risks as open-weight models—a compromised account, build pipeline, or storage system is sufficient. Reliable detection requires byte-by-byte comparison of weights against the original unmodified version, making trusted verification of the model's provenance critical.

Microsoft outlined its existing and planned defensive measures in the blog. Microsoft Foundry ties each model to its original publisher through source controls and runs models within Microsoft-owned containers, while simultaneously scanning the container environment for critical vulnerabilities. To address deeper threats—specifically cases where the trusted provider itself is compromised—Microsoft is developing weight-scanning tools powered by frontier AI models, red-teaming frameworks designed to detect anomalous model modifications, and cryptographic model signing mechanisms featuring verifiable provenance and tamper-evident audit trails.

출처:Microsoft Research

리서치

US AI Audio Firm ElevenLabs Launches Scribe v3 and v3 Live: First Realtime Speaker Diarization, Significant Error Rate Reductions

주제 · ElevenLabs중대
2026-11-10 20:00 GMT+8

US-based AI audio company ElevenLabs launched two new Speech to Text models on November 10, 2026: Scribe v3 for recorded audio and Scribe v3 Live for realtime audio. The company states that these models are designed to address challenges in real-world scenarios such as noisy environments, overlapping speakers, and mid-sentence language switching.

For batch processing, ElevenLabs reports that Scribe v3 reduces Word Error Rate (WER) by 27% compared to its predecessor, Scribe v2, across 102 languages. Improvements were observed in all tested languages, with German accuracy improving by 42%, Spanish by 35%, and English by 23%. Harder-to-transcribe languages like Swahili, Tamil, and Thai saw improvements ranging from 28% to 34%.

The realtime model, Scribe v3 Live, marks a key breakthrough. ElevenLabs claims it reduces semantic WER by 43% compared to Scribe v2 Realtime and introduces 'Speaker Diarization'—automatically labeling who said what—to realtime transcription for the first time. In recordings with three or more speakers, Scribe v3's diarization error rate is 17% lower than the previous generation. Additionally, improved 'Turn Detection' uses conversational context rather than just silence duration to determine when a user has finished speaking, preventing interruptions during mid-sentence pauses.

To meet enterprise needs, the new models offer enhanced output controls, including options to render numbers as digits or words, support for up to 1,000 keyterm prompts to improve proper noun recognition, and plain-text instructions for formatting rules. Scribe v3 Live also adds a multilingual mode supporting language switching within sentences.

Scribe v3 Live is currently available in the ElevenAgents platform, and both models can be accessed via the ElevenAPI. ElevenLabs notes that migrating from Scribe v2 requires changing only one line of code for the model ID. It is important to note that the performance data cited above is based on ElevenLabs' internal datasets and an adapted version of the FLEURS benchmark, representing vendor self-reported results that have not yet been independently verified.

출처:ElevenLabs

리서치

Hussman: The Three 'Free Money' Pillars of the U.S. Stock Market Bubble and the AI Profit Loop

주제 · Hussman중대
2026-11-11 00:51 GMT+8

This wave of speculation in the U.S. stock market has persisted for more than a decade. In his November 2026 market commentary, John P. Hussman, President of Hussman Investment Trust, offered a specific explanation: the bubble has not burst because it stands simultaneously on three pillars of 'free money.' The first is the Federal Reserve's zero-interest-rate expansion—from zero-interest Federal Reserve liabilities accounting for just 6% of GDP in 2008 swelling to 36% by early 2022, creating an unprecedented excess of liquidity in history. The second is the multi-trillion-dollar fiscal deficits during the pandemic, which directly subsidized corporate profits (through PPP loans) and subsequently indirectly subsidized profits as households spent down the savings they accumulated during the pandemic. The third, and most recent pillar, is the near-frantic infrastructure buildout in the field of artificial intelligence.

Hussman’s analysis of the third pillar is the most distinctive part of this commentary. He points out a simple arithmetic relationship: if corporate investment is surging, but the share of corporate free cash flow (profits minus net investment) relative to GDP has not risen, then profits must have exploded in tandem with investment. In other words, the sudden spike in corporate profits in the U.S. economy since 2022 directly depends on the frenzy of AI investment itself. This is not the narrative that 'AI improves productivity, so profits increase,' but rather a loop where 'corporations spend money on AI, and this expenditure itself becomes revenue for another corporation.'

He further uses the concept of 'dominance rents' to distinguish the nature of current profits. Hussman argues that the profit model of hyperscale cloud computing companies resembles what economists call a 'natural monopoly'—benefiting from declining marginal costs and economies of scale—rather than traditional monopolies that profit by restricting supply and raising prices. Sources of these profits include the private capture of network effects and the uncompensated use of publicly created content, which includes user-generated content on social media as well as materials used without compensation to train AI models. His judgment is that in equilibrium, if labor-saving technology depresses household income while corporate profits grow faster than the economy itself, the only way out is either for government deficits to expand further to bridge the gap, or for corporations to continuously reinvest profits into new net investment—which means the profits are undistributable.

At the valuation level, Hussman provides two specific figures. His most important indicator, MarketCap/GVA (the ratio of non-financial market capitalization to gross value added, including estimated foreign revenues), reached approximately 4.3 in early October 2026, while the historical median is about 1.1. Considering that current profit margins may be permanently higher than levels seen over the past century, he uses the GVA_PE10 metric adjusted by the ten-year average non-financial profit margin; the current level exceeds 25, whereas historically, levels associated with subsequent 12-year average S&P 500 returns of 10% were below 11. This yields a 'narrower baseline market risk estimate': a potential 55% downside when this market cycle completes. Without adjusting for profit margins, the potential downside is approximately 70%.

출처:Hussman Funds

리서치

James Picerno of The Capital Spectator: Which Signals Will Determine Whether This Rebound Sustains?

중대
2026-11-10 20:27 GMT+8

Monday’s rally brought relief to many investors, suggesting that last week’s sharp sell-off was merely another brief pullback—a pattern of scare-then-rebound that has repeated itself over the past few years. However, James Picerno, author of The Capital Spectator, warns that it is too early to determine whether the stock market has bottomed out. He lists several signals he is monitoring and admits that the evidence each signal currently provides is conditional.

The first notable technical fact is that the S&P 500 held its support at the 200-day moving average during this pullback. This widely followed trend line is generally regarded as a key reference for the market’s long-term direction and often becomes the focal point of battle between bulls and bears during periods of heightened volatility. Picerno believes that holding this level suggests investor sentiment remains constructive. He immediately adds an important caveat, however: relying solely on a rebound from the 200-day moving average to conclude that the stock market is ready to resume its upward trend is premature.

From a longer historical perspective, Picerno notes a pattern worth observing: since recovering from the year-long correction in 2022, the S&P 500 has experienced three steep but rapid drawdowns, each of which quickly reversed. He explicitly states that it is impossible to know whether this current decline will become the fourth, but the pattern appears to be repeating. He uses a rolling five-year window model to quantify this judgment, and the results show that the magnitude of this downturn is relatively mild compared to recent pullbacks. However, Picerno’s attitude toward this finding is cautious: a mild pullback may imply a correspondingly mild rebound, making range-bound oscillation the best-case scenario for the near term—a probabilistic assessment rather than a definitive conclusion.

At the sector level, the performance of technology stocks is another key observation point. The ratio of the Technology Select Sector SPDR Fund (XLK) to the S&P 500 SPDR Fund (SPY) has weakened significantly recently, which Picerno believes may signal that the AI-driven tech-led rally of recent years has temporarily come to an end. His conditional judgment is that if this ratio can stabilize and recover, the argument that the bull market still has room to run will be strengthened; conversely, if technology stocks fall further, it will pose a significant headwind for overall market sentiment.

On the macro front, the U.S. benchmark crude oil, West Texas Intermediate (WTI), fell to approximately $85 per barrel, its lowest level since late August last year. Picerno links the decline in oil prices to the possibility that tensions in the Iran conflict are cooling—U.S. President Donald Trump recently stated that the conflict is militarily effectively over—but he simultaneously points out that such statements require a healthy degree of skepticism. In his view, only sustained calm in the Persian Gulf will convince investors that the threat related to Iran is truly receding; moreover, how much war-related inflation remains to be transmitted through the global economy is still an unresolved question.

출처:capitalspectator.com

리서치

Linux Creator Says AI Fixes Outpace Bug Reports in BPF Subsystem

중대
2026-11-11 06:00 GMT+8

Linus Torvalds, the creator of Linux, stated in a mailing list announcement for the Linux 7.4-rc2 release that many of the recent changes were related to AI-generated fixes. He specifically noted that in the BPF (Berkeley Packet Filter) subsystem, the pace of AI-provided fixes has surpassed the rate at which new bug reports are arriving.

Torvalds described this situation as a sign that "the tide is beginning to turn," adding that if true, it is a welcome development. While he emphasized that other areas have not yet reached this stage and that some file changes were due to automated updates via Coccinelle, the tangible contribution of AI to core maintenance workflows is becoming significant.

Previously, the open-source community expressed concern that AI would flood projects with low-quality, hard-to-verify bug reports and patches, increasing maintainer burden. Torvalds' latest observation suggests that, at least in the complex BPF domain, AI-assisted coding is beginning to yield net positive productivity gains.

출처:テクノエッジ TechnoEdge

리서치

CoreWeave Reports $3.7B Q3 Revenue, Details Contract-Backed Financing Strategy

중대
2026-11-11 05:18 GMT+8

US-based AI cloud infrastructure company CoreWeave published a blog post on November 10, 2026, disclosing third-quarter financial performance and capital strategy. The company reported quarterly revenue of $3.7 billion, up 168% year-over-year, exceeding the high end of previous guidance, with more than $1 billion added quarter-over-quarter. Adjusted operating income reached $379 million, well above the guided range of $200 million to $260 million, with margins moving into the low double digits a quarter earlier than planned.

In terms of business scale, CoreWeave's revenue backlog grew to approximately $133 billion. Its global active power footprint increased to more than 1.7 gigawatts, doubling since the start of the year. The company attributed this expansion to new markets such as Indonesia and India, alongside continued growth in Denmark and Spain.

Regarding funding sources, CoreWeave detailed its financing model. The company stated that its investments are tied to committed demand, with loans for GPUs and other assets backed by contracted cash flows. This gives lenders confidence to lend against them, while customers typically help fund the buildout through prepayments. CoreWeave explicitly declared that it will not use the unsecured debt market to fund its business at current yields. This strategy aims to calibrate investment pace to customer demand, regulatory landscapes, and capital markets.

CoreWeave also noted that its platform has been unified under "CoreWeave Forge," enabling customers to build, run, and continuously improve AI in a single environment. The company claims that nine of the top ten AI foundation model providers run on its cloud, including AI natives like Wayve and Cognition, as well as traditional enterprises such as Caterpillar and Nissan.

출처:wf.coreweave.com

리서치

Former SignalFire Investor Sahir Raoof Raises $27.5 Million Debut Fund for AI Drug Development Startups

간단한 요점
2026-11-11 04:47 GMT+8

US venture capital firm Proximal Ventures has announced the close of its debut fund with $27.5 million. The firm was founded by Sahir Raoof, a former SignalFire investor who previously served as an executive at two healthcare startups. The fund targets startups that use artificial intelligence to transform the tools, infrastructure, and services supporting drug development.

Raoof argues that pharmaceutical companies have historically outsourced stages like molecular synthesis, manufacturing scaling, and clinical trial data analysis to specialized service providers. He believes AI can automate much of this work, allowing startups to deliver similar services at lower costs and significantly faster, thereby capturing billions in pharmaceutical industry spending. Proximal plans to invest in pre-seed and seed-stage startups, writing checks between $500,000 and $1.5 million.

The firm has already backed seven startups, including ChronicleBio, co-founded by former OpenAI executive Fidji Simo, which uses AI to analyze biological and medical data to understand chronic diseases and identify potential treatments. Proximal’s backers include VC firms SignalFire and Kleiner Perkins, Bertelsmann Healthcare Investments, and individual investors such as Dimension Capital co-founders Nan Li and Zavain Dar.

출처:TechCrunch

리서치

AI Security Firm Evolv Reports Q3 Results: Revenue Up 10%, Raises Full-Year EBITDA Guidance

주제 · Evolv AI安检公司간단한 요점
검증됨 2026-11-11 05:23 GMT+8

US-based AI security technology company Evolv Technologies Holdings, Inc. (NASDAQ: EVLV) released its financial results for the third quarter ended September 30, 2026, on November 10, 2026. The company reported total revenue of $47.3 million, a 10% increase year-over-year, while ending Annual Recurring Revenue (ARR) reached $140.5 million, growing 20% year-over-year. Although GAAP net loss widened to $15.6 million, Adjusted EBITDA remained positive at $4.7 million.

Looking ahead, Evolv narrowed and raised the lower end of its full-year 2026 revenue guidance from the previously issued $180–$185 million range to $182–$185 million. The company also increased its expected full-year Adjusted EBITDA from $15–$16 million to $15.5–$16.5 million. Management noted that despite a challenging prior-year comparison that included the largest contract in the company's history, the quarter achieved double-digit revenue growth with ARR expanding faster than revenue, reflecting progress toward a more predictable subscription model.

출처:sec.gov

리서치

CoreWeave Launches 'Zero Trust AI Framework' Integrating Key Control and Confidential Computing for Enterprise Security

중대
2026-10-21 19:30 GMT+8

US-based AI cloud infrastructure company CoreWeave announced the "CoreWeave Zero Trust AI Framework™" on October 21, 2026. The framework aims to address the core trust issue enterprises face when migrating sensitive data to the cloud for AI training and inference: whether the cloud provider retains access to customer data and models. CoreWeave states that the framework shifts security responsibility from "relying on vendor policies" to "technical boundaries controlled by the customer" through six layers of protection.

Previously, CoreWeave launched Remote Key Encryption (RKE) in September, allowing customers to retain encryption keys in their own key management systems, ensuring CoreWeave holds only ciphertext. This new framework expands on that foundation with additional dimensions, including Support Access Management (SAM), hardware root-of-trust attestation, and planned confidential computing support for NVIDIA Blackwell and Vera Rubin platforms. Notably, SAM allows customers to decide when CoreWeave engineers are permitted to enter the infrastructure for troubleshooting, breaking the traditional cloud service model where providers have default privileged access.

It is important to note that some components remain in planning or early stages. In its blog post, CoreWeave explicitly uses the phrase "intends to support" to describe its confidential computing capabilities on NVIDIA Blackwell and Vera Rubin platforms, indicating that while the overall architecture is announced, certain advanced security features are not yet fully available across all environments. Dave McCarthy, Group Vice President at IDC, commented that this move represents a critical step in shifting trust from vendor promises to technical controls, which is essential for enterprises deploying proprietary AI at scale.

출처:wf.coreweave.com

리서치
다음 →
로그인 / 회원가입
독서리서치레이더투자 프레임워크

Nevertheless, Rasmussen's remarks are noteworthy because they represent a forming direction in enterprise practice: the authority to judge the economics of AI agents is shifting from IT departments to business leaders. Her own approach is to first map all roles—'What skill directions will future business analyst roles migrate to? Do the people in these positions have the ability to become forward-deployed engineers?'—and then use this map to talk with employees, telling them where the direction lies and what needs development, rather than announcing job eliminations after the fact.

She simultaneously acknowledged this might be 'one of the hardest conversations to have.' Over the past two years, Dayforce has done extensive change management—AI marathons, external speakers, open tools—which Rasmussen believes is part of what the company should do, but she also returns ownership to employees: 'You cannot wait for the company to take you where you need to go. Especially for technical roles, you must use personal time to pull yourself up to speed.' She added that this does not mean learning only outside of work; there are internal channels too, but 'now is really the time to ensure you are not falling behind.' This statement places the responsibility allocation of AI transformation on the table: the company provides tools and direction, and employees provide adaptation speed.

Papailiopoulos emphasized in his article that this is not an argument against open-weight models, but rather a universal software supply chain security issue: model weights should be audited just like code. He also pointed out that no scan can absolutely prove a model is safe; this should be treated as one layer within a defense-in-depth strategy, rather than the sole solution.

However, Hussman’s investment operations are not simply a matter of 'shorting and waiting.' He admits that since gaining an insight into 'non-duality' earlier this year—that they had previously been 'discriminating against the bubbles'—their investment stance has become more flexible. At extreme valuations, all operations must have a safety net, but when short-term market behavior interacts with measurable investor psychological factors, they are willing to take a constructive tilt about 20% of the time. He gives a specific portfolio optimization example: if investors allocate between the S&P 500 and cash in a 60/40 configuration, then over the past 1, 3, 5, 7, and 10 years, the lowest-risk allocation matching the return of a traditional 60/40 portfolio typically assigns 33-42% to their strategy, with the remainder going to the S&P 500.

The most noteworthy aspect of this commentary for readers is not the 55-70% figure itself—any long-term bear would offer a similar estimate—but Hussman’s structural deconstruction of the source of profits. He redefines the phenomenon of 'surging corporate profits,' usually seen as fundamental support, as fuel for the bubble itself: AI capital expenditure creates profits, profits provide justification for more AI capital expenditure, and the entire loop depends on the continuous supply of 'free money' (zero interest rates, deficits, and investors willing to throw money at new technologies). His core warning is: do not confuse the sound of money being spent with the sound of wealth being created.

The greatest uncertainty Hussman himself acknowledges is timing. He states that 'the greatest investment risk at the moment is the inclination to confuse the length of time a bubble has persisted with proof that a bubble does not exist.' His valuation framework provides expectations for returns over ten to twelve years, rather than timing signals. He explicitly clarifies that their investment discipline does not rely on future market declines or economic disruptions, nor does it presuppose when disappointments will arise. This is an argument about structural fragility, not a prediction of trigger points.

Picerno’s final judgment is conditional: whether this rebound marks the start of a lasting recovery or merely a temporary respite depends on whether the market can simultaneously maintain technical support, withstand geopolitical risks, and preserve confidence in the broader economic outlook. He does not assign weights or joint probabilities to these signals, nor does he declare any single indicator to be decisive. The value of these observational indicators lies in providing investors with specific tracking anchors to help accumulate judgment basis gradually amidst uncertainty—rather than offering one-time definitive answers.