ReadingResearchRadarInvestment framework
Sign in / Sign up
AllTechnologyProductsInvestingPublic marketsOther

Reading

2026-10-0832 posts

Refresh

Robot Data Startup Mecka Raises $60 Million Series B Led by Sequoia

Topic · Mecka AI融资动态Material
Verified 2026-10-08 08:16 GMT+8

Mecka AI announced the completion of a $60 million Series B funding round, led by Sequoia Capital with participation from Nvidia and Microsoft's venture fund M12.

Founded in 2024, the startup pays individuals wearing body sensors to record everyday tasks like making coffee or fixing cars. This generates high-quality human motion data for training humanoid robots, mirroring how Scale AI provides labeled data for large language models.

Prior reporting indicated Mecka was negotiating this round at a $500 million valuation. The new capital will accelerate its expansion in embodied intelligence data infrastructure amid competition from firms like XDOF.

Sources:mecka.ai

Research

Open-source AI Firm Nous Raises $90 Million at a $1.5 Billion Valuation

Material
2026-10-07 23:00 GMT+8

Nous Research, the developer of the open-source Hermes agent, has confirmed a $90 million Series B raise at a $1.5 billion valuation, bringing its total funding to $158 million.

The round was led by Robot Ventures with participation from Nvidia, Samsung, Union Square Ventures, and others. The capital will fund the launch of "Hermes for Businesses," allowing companies to deploy customized AI agents for multi-step workflows while keeping data private.

According to TechCrunch, the open-source Hermes agent has been cloned over 24 million times, driving roughly 2.5% of global AI token usage (a company estimate). The WSJ previously reported that Nous had approximately $36 million in annualized revenue by mid-September 2026 and expects to pass $100 million before the end of the year.

Sources:nousresearch.com

Research

Venture Fund Endeavor Catalyst Raises Capital for Founders Outside Major Tech Hubs

Material
2026-10-08 06:59 GMT+8

Endeavor Catalyst has closed its fifth fund with $320 million in capital commitments, bringing the firm's total assets under management to more than $850 million.

As the venture arm of the New York-based nonprofit Endeavor, the fund specifically backs founders outside major tech hubs. Approximately 90% of its investments are outside the U.S., with Europe identified as its fastest-growing region. Unlike traditional VCs, Catalyst typically joins rounds led by other institutional investors at similar terms, writing checks usually between $1 million and $3 million.

Linda Rottenberg, co-founder of Endeavor, stated that half of the fund's profits go back to the nonprofit to support the next generation of founders building "elsewhere." This structure aims to counter the current trend of venture capital concentrating heavily on AI companies in San Francisco.

Sources:techcrunch.com

Research

First Prison Sentence for AI Music Fraud

Topic · AI音乐流媒体欺诈案Material
2026-10-08 20:19 GMT+8

Michael Smith, a North Carolina resident, was sentenced to 18 months in prison and fined $8 million for using automated bots to stream his own AI-generated songs and defraud royalty payments.

This marks the first-ever criminal case involving AI-assisted music streaming fraud. Between 2017 and 2024, Smith created hundreds of thousands of AI tracks and operated up to 10,000 bots on platforms like Spotify and YouTube Music. The DOJ noted that in April 2023 alone, his bot accounts generated 80.9 million streams on YouTube Music, nearly nine times the volume of Taylor Swift's entire catalog during the same period.

The ruling establishes criminal liability for combining AI-generated content with automation for financial gain, serving as a warning to streaming platforms regarding anti-fraud measures and AI content compliance.

Sources:tomshardware.com

Research

Ethereum Researchers Debate AI Threat to Wallets

Material
2026-10-08 21:52 GMT+8

Leading Ethereum researchers Justin Drake and Vitalik Buterin are publicly debating the risk that AI-assisted mathematics could break ECDSA signature schemes.

Drake called for a "bunker mode" on X, urging users to move funds to addresses that have never signed a transaction. Signing reveals the public key, which attackers could theoretically use with AI to derive the private key; addresses showing only the public key hash remain protected.

Buterin agreed on the need for a hash-function-based architecture long-term but warned against acting too fast. He stated he has lost more money to botched migrations than to hacks. No one has actually broken ECDSA in practice yet, making this dispute primarily about the perceived timeline of the threat and the cost-benefit analysis of defensive strategies.

Sources:the-decoder.com

Research

Google's Sashiko AI Reviews 191k Kernel Patches

Material
Verified 2026-10-08 19:55 GMT+8

Google engineer Roman Gushchin disclosed at the Linux Plumbers Conference on October 6 that Sashiko, an AI-powered code review tool, has completed more than 191,000 patch reviews in under a year.

The tool covered 99 mailing lists, performed over 19 million Git lookups, and received 7,600 replies from 1,100+ kernel developers. This year, 463 Linux kernel CVEs have cited Sashiko, along with 1,277 upstream commits and 1,567 linux-next commits.

These figures come from conference presentation materials; the specific criteria for “citation” and details on independent audits were not specified in the public summary.

Sources:lpc.events

Research

Fewer parameters, faster training, but don't treat it as a universal solution when switching tasks

Material
Verified 2026-10-08 00:03 GMT+8

Long-term reading · 《ALBERT: A Lite BERT for Self-supervised Learning of Language Representations》(2019)

ALBERT is an approach to saving parameters in text models: it reduces the size of word representations and shares parameters across layers, enabling faster learning of sentence meanings with fewer parameters. When you see it achieving high scores on text tasks with fewer parameters, first ask where those saved parameters come from.

The mechanism it provides is this: parameters are saved by reducing the size of word representations and sharing them across layers, while using sentence-order prediction to retain multi-sentence task capabilities. Therefore, when reading about low-parameter text models, ask whether these two aspects still fit your specific task.

If dealing with non-English languages, extremely small datasets, or if the width per layer is too large, do not use it to judge general optimality. In some single-sentence tasks, removing the sentence-order objective has limited impact.

"ALBERT" (2019) | Next review date: 2027-09-20

Sources:《ALBERT: A Lite BERT for Self-supervised Learning of Language Representations》(2019)

Research

"This Time Is Different": Why Debt Risk Cannot Be Judged by Totals Alone

Verified 2026-10-08 22:04 GMT+8

Long-term reading · 《This Time Is Different》(2009)

"This Time Is Different" does not discuss when the next crisis will arrive, but rather why investors always believe that old risks have disappeared during periods of prosperity. Reinhart and Rogoff examine financial crises across different countries and eras together: two waves of default can be separated by many years, and calm itself easily leads people to mistakenly believe that institutions have changed.

The first concept worth understanding is "debt intolerance": how much debt a country can bear cannot be judged solely by a uniform ratio. Chapter 2 of the book links weak institutions and difficult political decision-making with repeated external borrowing. Borrowing temporarily postpones the difficult choices of adjusting taxes and spending, but does not solve the problems behind these choices; while debt may appear still rollable, the basis for repayment may already have weakened.

The second concept is "seeing the complete debt picture." Chapter 8 explains that some countries default even when external debt appears low, partly because researchers overlook domestic debt. When domestic burdens are included, fiscal pressure no longer seems minor. Therefore, a small scale of external debt alone cannot prove safety; one must ask what other debts the same fiscal revenue must also service.

Applying this framework to today's sovereign debt research involves separating two questions: whether borrowing has improved future repayment capacity, and whether currently visible debt covers all obligations. This provides a method for reading financial and institutional materials; it is not possible to judge whether a specific country is currently dangerous based solely on historical samples from the book.

For example, suppose a country has low external debt but heavy domestic debt, and borrowed funds are primarily used to maintain expenditures. Looking only at external debt would misjudge it as having a low burden; looking only at recent absence of defaults might ignore that borrowing has been substituting for difficult fiscal adjustments. Conversely, if the complete debt picture and sources of repayment are verified, risk comparisons can be made more soundly. This scenario illustrates the mechanism and is not a judgment about any real-world country.

This book is suitable for establishing historical references for risk judgment, but not as a countdown to crisis. Institutions, financing structures, and response capabilities differ between countries, and patterns repeatedly appearing in history do not mean the next instance will necessarily follow them exactly. This article is written based on publicly available excerpts from the original book; full access to the entire text was not obtained.

Recommended reading: "This Time Is Different: Eight Centuries of Financial Folly" (2009) by Carmen Reinhart and Kenneth Rogoff. Readers interested in sovereign credit or macro risks should first read Chapter 2 "Debt Intolerance" and Chapter 8 "Domestic Debt," then read the whole book keeping in mind the two questions: "Has the repayment base improved? Are any burdens being overlooked?"

Sources:《This Time Is Different》(2009)

Research

You’re all caught up in this view

Sign in / Sign up
ReadingResearchRadarInvestment framework