这个案例展示了AI系统在专业领域的应用能力，20万美元的API成本反映了高质量AI评估的显著经济成本。99%的GCC torture test通过率是一个令人印象深刻的指标，表明AI系统在特定领域可以达到接近人类专家的水平。

这个案例展示了开放世界评估的实际应用，每年约5万美元的成本表明这种评估需要相当大的资源投入。相比传统基准测试，这种评估方式更接近真实应用场景，但也因此成本更高，难以大规模实施。

AI硬件成本占AI公司总支出的一半以上(54%-62%)，这凸显了计算资源在AI开发中的核心地位。如此高的比例表明，AI公司的竞争很大程度上转化为对计算资源的获取和利用能力的竞争。这也解释了为什么各大公司愿意为芯片支付高价并积极投资自研芯片。

令人惊讶的是：Anthropic声称每次代码库扫描要花费2万美元来发现零日漏洞，这个价格远高于人们对于代码安全审计成本的预期，揭示了AI安全分析的高昂门槛。

运行一次 452 个任务的评测，GPT-5.4 花费 1110 美元，Claude Opus 4.6 花费 1055 美元——每个任务平均约 2.3 美元。而 Gemini 3 Flash 只需要 596 美元，实现了 27.7% 的成绩（vs 顶级模型的 33.3%）。这个性价比数据对 AI 选型决策极为关键：如果业务场景可以接受 27% 而非 33% 的成功率，Gemini 3 Flash 能节省近一半成本。在金融服务的大规模部署中，这个差异将被放大数千倍。

大多数人认为AI视频分析仍处于早期阶段且成本高昂，但作者指出AI视频分析成本已大幅下降40倍，质量反而提升。这一反直觉观点暗示视频分析可能已经跨越了实用性的门槛，将催生全新的应用类别，挑战了人们对AI视频处理能力的传统认知。

Market analysis of library card catalogs in 2023.

As card catalogs lost their functionality in libraries and were de-acquisitioned there was a wave of nostalgia which caused people to purchase them, often in auctions, at higher than expected prices. Once they had them, most of these purchasers realized that they didn't have functional uses in their homes for them (beyond wine or liquor bottle storage, small crafts, or use as a zettelkasten, which seem to be the only reasonable upcycling use cases I've seen and the last seems to be very rare and niche). They sit and take up space for very little value in return beyond some esthetic beauty and nostalgia. As a result many soured on their ownership. Most owners naturally want to recoup their original purchase price thinking that relative rarity will save them.

Combined with this there was a resurgence in mid-century design esthetic which had some furniture restorers and designers buying and doing full (and very pretty) expensive restorations of older 20s - 40s versions which sold at auctions for $4,500 and up. Given the rarity of some of these older, fine furniture versions along with the work in restoration and the limited market only those who had a tinge of nostalgia and money to burn made purchases which resulted in a limited number of actual sales.

These two factors mean that almost all of the listings for library card catalogs are heavily overvalued on eBay, Facebook Marketplace, Craig's List, Etsy, etc. The fine furniture restorations have set an artificially high price point which some feel theirs must match as well. The difference in quality however is stark. Because of their size and lack of functionality, there is a relative glut of them on the market which all bear inflated prices. Those who originally spent inordinate amounts for them, feel they will still have that same value to others, so they list them online for inflated prices.

I've been closely watching the online "market" for them for over a year and see the same several dozen or more listed across the country usually in the range of about $30-$60 per drawer. Many are listed as local pick up only, which further hampers the overall market. This also brings up the issue of shipping a 60 drawer card catalog which can easily run in the $800-$1,500+ range which usually requires additional shipping logistics involved with freight. Most catalogs are already overpriced, but adding an additional $1000 tax on top is a bridge too far for all but the highest end of the market. Some platforms like Etsy and eBay which take cuts of the final sale also add to the cost of the sale.

In the year and a half or more that I've been watching, I've only seen a handful of actual sales, all of which were local, and many of which were in the Los Angeles area. All of these sales have been for listings which eventually were reduced down to the $15 per drawer range. One local sale was in Wisconsin was for $10 per drawer (a 30 drawer file) and another in Los Angeles was for $12.50 per drawer (on a 20 drawer file).

A note on condition

Outside of a small handful of fine furniture listings in the $4,000+ range, most ex-Library card catalogs are generally very well worn and not in great condition which makes them less valuable as decoration pieces. In fact, many are often missing their original card catalog rods, have dents, dings, or other cosmetic issues. Some are missing drawers or have replacement drawers which don't match. Some may be slightly mismatched having been purchased in different eras as modular pieces and put together. Frequently they have been modified from their original states to include inserts or other material to fill in the holes which where almost standard in the bottoms of the drawers.

Advice

If you're in the market, know that it is tremendously inflated, a fact which most sellers are aware of as they've got them listed, some for many years, not resulting in actual sales. If you really want one and find it in a reasonable condition, I highly recommend making an offer for it at about $10 per drawer and potentially go up to $15. Anything higher than that is overpaying based on actual recent market conditions. If you have the money to burn, feel free, but keep in mind that like many others in the past, once the initial nostalgia has passed, you've probably got a large piece of relatively non-functional furniture in your home.

It's not common, but some government auction sites will list card catalogs for auction from time to time. Because they actively want to sell them these can be purchased in the $2-10 per drawer range or less. Often they tend toward the larger 60+ drawer range, aren't in good condition, or need to be picked up and shipped to your final destination, usually within a few days of purchase as the original owners don't or explicitly won't handle shipping. These are likely to need some restoration work to be decorative pieces in many homes.

If you want something brand new, you can check out Brodart, which is the only remaining card catalog manufacturer/sales firm I'm aware of in the United States. Their systems are modular, so you can pick and choose what you'd like to have. The only caveat is that they start at $1,700 for their smallest 9 drawer model and can go up to $11,648 (plus shipping) for a full 60 drawer model. The other potential drawback, for some, is that they are made of a mixture of wood, metal and plastic versus the all wood and metal fittings of older vintage models.

If you're in the market primarily for nostalgic reasons, then you might also consider looking at some of the older desktop wooden card catalogs which are often much less expensive, take up far less space, and can be wonderfully decorative. Some of the smaller two to six drawer desktop models have the benefit of potentially serving as recipe boxes or paper rolodexes, zettelkasten, or simply small office storage. Here again, the online markets are likely to be heavily overpriced with 2 drawer models being continually listed at $150 and 4 drawer models in the $250-400 range. These sellers know that these prices don't result in actual sales as they've been sitting on them for long periods of time (presumably hoping to get lucky). Here I'd recommend you make offers in the $20-30 per drawer range to see what you can find. Another benefit is that these smaller models are far cheaper to ship across the country. For additional advice on these, see: The Ultimate Guide to Zettelkasten Index Card Storage.

Study: Myocarditis risk 37 times higher for children with COVID-19 than uninfected peers | American Academy of Pediatrics. (n.d.). Retrieved October 10, 2021, from https://www.aappublications.org/news/2021/08/31/covid-myocarditis-risk-children-083121

Zhou, F., Shefer, A., Wenger, J., Messonnier, M., Wang, L. Y., Lopez, A., Moore, M., Murphy, T. V., Cortese, M., & Rodewald, L. (2014). Economic Evaluation of the Routine Childhood Immunization Program in the United States, 2009. Pediatrics, 133(4), 577–585. https://doi.org/10.1542/peds.2013-0698

ReconfigBehSci on Twitter. (n.d.). Twitter. Retrieved 3 March 2021, from https://twitter.com/SciBeh/status/1351453660396605440

Graham, B. S. (2020). Rapid COVID-19 vaccine development. Science, 368(6494), 945–946. https://doi.org/10.1126/science.abb8923

Gupta, Prateek, Tegan Maharaj, Martin Weiss, Nasim Rahaman, Hannah Alsdurf, Abhinav Sharma, Nanor Minoyan, et al. ‘COVI-AgentSim: An Agent-Based Model for Evaluating Methods of Digital Contact Tracing’. ArXiv:2010.16004 [Cs], 29 October 2020. http://arxiv.org/abs/2010.16004.

McLaughlin, Patrick A, and Casey B Mulligan. ‘Three Myths about Federal Regulation’. Working Paper. Working Paper Series. National Bureau of Economic Research, May 2020. https://doi.org/10.3386/w27233.

Feys, F., Brokken, S., & De Peuter, S. (2020, May 22). Risk-benefit and cost-utility analysis for COVID-19 lockdown in Belgium: the impact on mental health and wellbeing. https://doi.org/10.31234/osf.io/xczb3

Verbruggen, R. (2020 March 24). Another COVID Cost-Benefit Analysis. National Review. https://www.nationalreview.com/corner/another-covid-cost-benefit-analysis/

Estimated economic benefit of data linkage

the potential value from linking Census data to administrative data sets is only beginning to be realised and holds immense potential.(In other work for the Population Health Research Network, Lateral Economics concluded that data linkage generated over $16 for every dollar invested).

Economic benefit of the Census.

Our estimates suggest the benefits of running the Census easily outweigh its costs in the order of$6 of economic value for each $1 it costs. On this reckoning, the cost of the Census would have to rise to six times its current cost –to around $3 billion every five years –before it startedto become cost ineffective