Glass-based substrates are slowly beginning to push out organic substrates commonly used in PCBs due to often superior material properties. One area where glass substrates have however struggled is with through-hole vias and providing the conductive copper path through them. A 2024 article by [Keith Best] gives a good overview of the topic, with recent news showing how much companies like Intel are pushing for glass substrates, specifically for the packaging of dies.

One major advantage with vias in glass substrates is that they can be much smaller, enabling smaller than 0.1 mm diameter holes with far finer pitch. The challenge here is to make perfect holes with a laser that are defect-free, as well as have the intended diameter.

After that this through-glass via (TGV) has to be coated or filled with copper, much like their organic equivalent. Said TGV can be fully filled with copper, or use plating and add dielectric filler. Detecting flaws in such a finished TGV is important.

In a 2025 review article of glass substrate technologies by [Pratik Nimbalkar] et al. published in Chips the state of the art at the time was covered. The need for ever higher-density integration options with ASICs is highlight here, especially now that many chips today consist of multiple interconnected dies inside a single package.

The complications of creating TGVs with femtosecond laser pulses in Borofloat 33 glass are highlighted by [Daniel Franz] et al. in a 2025 research article, with microcracks and backside ablation observed without proper precautions, something which previously was often resolved by an etching step following said laser drilling. The main issue here is the post-drilling residual stress from the thermal shock, which the authors demonstrate can be largely prevented with careful tweaking of the laser drilling parameters.

As pointed out in a 2024 review article by [Chen Yu] et al. glass substrates are useful for far more than just high-density chip packaging. Glass substrates are also chemically resistant, have a higher heat resistance, are largely transparent to RF and can be hermetically sealed against outside influences. This makes them great for various advanced sensors and communication devices.

Meanwhile, if you wanted to do some metal-depositing on glass at home, we covered this recently.

The PlayStation Portable, or the PSP, was discontinued over a decade ago, but its cultural presence never fully faded. Most recently, the fast-fashion brand Zara gave it a second life in the most unexpected form: the company has dropped a crossbody bag modelled after the PSP 1000. 

The Zara PSP Crossbody Bag is exactly as delightful and absurd as the name sounds. The shadow drop came without an announcements or media campaigns, but the retro gaming community has already taken notice of it. 

What does the bag actually look like?

It is actually a relatively small crossbody bag whose front face is a silicone recreation of the PSP 1000, in convincing detail. The bag comes with embossed buttons, logos (on the front and the back), an analogue nub, and a vinyl panel standing in for the iconic 4.3-inch widescreen display. 

The adjustable shoulder strap also carries a PSP branding, along with the classic triangle, circle, cross, and square shapes. Clearly, Zara doesn’t want the product to look like a cheap knockoff, and the result shows. 

The bag measures 4.3 x 7.9 x 2 inches, has a main zipper compartment, and is made from polyurethane thermoplastic on the front face with a silicone overlay and a polyester shell and lining.

How much does it cost, and where can you get it?

For now, the Zara PSP Crossbody Bag is available at $35.90 in the United States and £19.99 in the United Kingdom, available directly via the company’s official website and in stores. The bag is only available in one color, black.

I also see the trademark symbol on the website, implying that this is some sort of licensed deal between Zara and Sony, rather than an unofficial product, even though neither company has confirmed the arrangement. 

At $35.90, it could be among the most affordable pieces of PSP memorabilia you might ever own, but only if the PSP mattered to you. 

When you think of Memorial Day sales, you probably think of mattresses and other home goods. And while those items are definitely discounted, now is also a good time to purchase tech. Personally, I’m not buying anything right now unless it’s discounted—and fortunately many of our top picks are. Whether you’re shopping for a power bank, a new pair of headphones, or some other gadget, I’ve rounded up the best Memorial Day deals for your perusal. Most of these deals end at the end of the day.

Check out our buying guides for more recommendations, including the best headphones, the best laptops, and the best cheap phones. You might also want to check out our additional Memorial Day deals coverage.

Updated Monday, May 25: We’ve checked prices, removed expired deals, added 6 new deals, and ensured accuracy throughout.

WIRED Featured Deals:

Sony WH-1000XM5 for $248 ($152 off)

The Sony WH-1000XM5 have a very frustrating name, but they’re the predecessor to our favorite wireless headphones, and they’re still an excellent pick if you don’t want to shell out for the new WH-1000XM6. They go on sale frequently, but rarely drop this low in price, which comes within $5 of their all-time low. If you’re in the market for over-ear headphones, they’re hard to beat. They’re comfortable, portable, lightweight, and stylish, and they’ll make your music sound great no matter what you like to listen to.

Diotima received €500,000 under Enterprise Ireland’s Commercialisation Fund last year.

AI edtech start-up Diotima, founded by former secondary school teacher Siobhan Ryan, has spun out from Trinity College Dublin (TCD).

The platform aims to enable educators to use AI to create assessments and individualised feedback to improve learning outcomes and lighten burdens on teachers.

The spin-out will be led by edtech commercialisation specialist Jonathan Dempsey as CEO, with Ryan, also a biochemist and environmental scientist, becoming chief product officer and learning lead.

Development engineer Daniel Fernandez and AI engineer Dr Long Mai, who have both worked on the Diotima project, will also join the inaugural team.

Dr Eoin Lane, an AI regulatory compliance expert who was formerly the global head of AI and data science at the Bank of New York Mellon, is a governance consultant to the Diotima project.

“This all started when I was working as a teacher and I had a vision for how AI could enhance teaching and learning even before any of the models like ChatGPT launched,” said Ryan.

“I then worked with Tom Pollock and Learnovate to develop this vision into a real-world project.”

Diotima began its partnership with Learnovate in February 2025 and received €500,000 in funding under Enterprise Ireland’s Commercialisation Fund, which supports third-level researchers in translating their research into commercially viable solutions.

The idea was to develop an AI-enabled edtech platform to help teachers and other educators create assessments, as well as provide feedback to learners, all in compliance with European and Irish legislation.

Specifically, the platform meets requirements under the EU AI Act, which has strict regulations around the usage of AI in high-risk sectors such as education.

“We aim to position Diotima as a leader in responsible AI for education,” Ryan said. Diotima will continue to engage with prospective customers and stakeholders for a go-to-market strategy while also seeking new investment.

“Using responsible AI, Diotima promises to develop into a revolutionary platform for learners in secondary schools and professional education organisations, delivering formative feedback and better outcomes overall,” said Pollock, Learnovate’s impact, licensing and commercialisation manager.

Learnovate launched its ‘Responsible AI for Learning’ initiative earlier this year to enable AI implementers and practitioners involved in teaching and learning to share knowledge, interpret guidelines and comply with AI regulations.

The initiative is made up of professionals from all four education domains – schools, higher education, vocational education and training, and professional education – as well as representatives from the Department of Education, teaching unions and other sectors.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

• AI-powered hackers now exploit software flaws faster than companies can patch systems
• Mobile phishing scams now outperform traditional email attacks across corporate environments worldwide
• Unauthorized AI tools are quietly leaking sensitive company information across global workplaces

For the first time in nearly two decades, exploiting software vulnerabilities has overtaken stolen passwords as the primary way hackers breach corporate networks.

Verizon’s 2026 Data Breach Investigations Report claims the exploitation of vulnerabilities now accounts for 31% of all confirmed data breaches.

Over the past two decades, technical debt meant outdated architecture, messy code, and poorly maintained documentation. That definition is no longer sufficient in the AI era, where failure modes are more subtle and often non-linear. AI systems are introducing new layers of technical debt that live across prompts, models, and data dependencies — making these layers less visible, harder to measure, and often more dangerous than traditional debt.

A crisis hiding in plain sight

The complexities of AI systems and their associated failures have been well documented. A 2025 MIT study found that 95% of AI projects fail to reach production or deliver value. A similar study by S&P Global Market Intelligence found that 42% of businesses scrapped multiple AI initiatives in 2025 — a sharp increase from 17% the previous year. Various reasons are cited for these failures, but most of them point to poorly designed and implemented systems that are complex to manage and have multiple hard-to-monitor failure points, leading to a rapid accumulation of AI debt. 

Traditional technical debt was localized to the codebase, and bugs were usually easily reproducible. Consequently, bugs could be easily identified during tests and fixed through rearchitecting the codebase. However, AI debt is much more distributed, manifesting across prompts, models, data pipelines, and all associated infrastructure. It is also more intermittent: Due to the probabilistic nature of AI, systems do not always respond the same way, leading to intermittent failures. This makes it much more challenging to identify risks during testing, and also creates a need for more continuous monitoring even post-deployment to prevent gradual drift and worsening performance.

The new forms of AI debt

AI debt typically manifests across four new forms, each of which comes with its own set of risks.

Prompt debt is the most visible of these. A modern version of ‘spaghetti code,’ this can include undocumented prompt tweaks, accumulated ‘quick-fix’ prompts that lead to inconsistencies, neglected version control of prompts, and ‘prompt stuffing’ (the cramming of extraneous data or context directly into AI prompts). All these combine to make prompts a form of untyped, untested code without any version control, leading to increased brittleness and vulnerabilities.

Model dependency debt is another increasingly common form of AI debt. Most enterprises now depend on a mixture of external models developed by leading foundation model providers; applications and agents are built on top of API calls to these models. Consequently, application logic now depends on models that are external to the core system, and that cannot be clearly controlled. As models update, performance varies and reproducibility is lost — prompts tuned for one model may fail or perform poorly when switched to another model, whether an update from the same provider or from another provider.

Most enterprise AI deployments today use retrieval-augmented generation (RAG), which pulls in additional context from enterprise data repositories. Retrieval debt is a consequence of these repositories having messy data, duplicated documents, and outdated information. This causes AI to return technically correct answers that are outdated and no longer relevant, causing downstream failures. Unlike hallucinations, these are harder to detect because they were correct, perhaps even until recently, and hence look correct to any tester. 

Evaluation debt reflects the lack of standardization in testing and monitoring for AI models and applications. While AI benchmarks exist, they tend to focus on narrow tests and reflect point-in-time results. Most enterprises lack consistent testing standards, ground truth datasets, and real-time monitoring of deployments; there is no equivalent yet of continuous integration /continuous delivery (CI/CD) for prompts. As a consequence, CIOs and CTOs do not have clear visibility into model performance and cannot track improvements or worsening of models. 

All of these are in addition to traditional forms of technical debt, which still manifest across the tools and systems that AI applications and agents interact with, read from, or write to. A rapid increase in the adoption of AI-generated code (often deployed without inadequate testing) is further aggravating inconsistencies within, and poor maintainability of traditional codebases. 

The new forms of AI debt combine with these earlier forms of technical debt to compound rapidly and create large-scale risks that can cause catastrophic failure of entire enterprise deployments. Solving for these risks is made even more challenging by the distributed nature of AI ownership – most systems span engineering, product, data, and business teams, leading to unclear accountability when an error is identified. 

As a result, these risks manifest in the form of escalating compute costs, inaccuracies in AI outputs, and increasing exceptions that need to be handled by humans — leading to projects often stalling and failing due to unclear return-on-investment stories and a lack of trust from users. 

How enterprises can prevent AI debt

AI debt will not be solved by ‘better’ models — failure rates remain high despite models already having high accuracy. The solution to AI debt requires better system design, integration, controls, and changes in organizational culture. 

First, prompts need to be treated as code. This involves careful version control, documentation, and rigorous testing both pre- and post-deployment for all possible prompt configurations. Best practices from the traditional world of coding — such as the use of smaller prompt blocks instead of large prompt-stuffed walls, or reducing the use of hard-coded parameters — can also help mitigate AI debt. 

Second, evaluation needs to be built into the entire AI infrastructure stack. Continuous evaluation pipelines need to be established and must reflect a wide variety of metrics measuring both technical and business-aligned metrics. In addition, AI observability systems should be integrated to monitor output quality, failure rates, model drift, and data drift.

Third, explainability should be included by default in all AI results to make up for limited reproducibility. Data lineage, models used, and the steps followed should be clearly traceable so as to allow auditability of results and correction in case of any systemic errors. 

This requires explicit AI debt reduction programs and associated budgets, similar to earlier waves of investment in security or in cloud modernization. These need to be driven at a CXO level by key leaders to prevent costly rework later.

Conclusion: A stitch in time

Enterprise AI deployments are not just static code; they are living systems that interact with the entire enterprise stack. As a result, the defining challenge in an agentic enterprise will not be building or deploying intelligent systems, it will be maintaining these systems to ensure continued reliability during real-world operation.

Enterprises that seek to proactively identify and mitigate AI debt from the design phase itself are the likeliest to build sustainable AI platforms that deliver significant long-term productivity boosts across the organization. 

Vikram is a principal at Cota Capital, where he invests in early-stage enterprise tech and deep tech companies.

Welcome to the VentureBeat community!

Our guest posting program is where technical experts share insights and provide neutral, non-vested deep dives on AI, data infrastructure, cybersecurity and other cutting-edge technologies shaping the future of enterprise.

Read more from our guest post program — and check out our guidelines if you’re interested in contributing an article of your own!

• Kansas City schools replace 30,000 Windows and Chromebooks with Apple devices
• Concerns raised over financial loss from retiring functional school computers
• District cites security, durability, and “student pride” as reasons for the Apple switch

The Kansas City Public Schools district has announced a sweeping transition which will remove tens of thousands of non-Apple devices from its classrooms.

According to information on the district’s website, administrators will replace more than 30,000 Windows PCs and Chromebooks with Apple hardware over the coming months.