My dad’s old-school bookkeeping habit

For as long as I can remember, my father has run a small business. And for just as long, he has kept his books the old-fashioned way: every sale gets written down by hand. It works well enough for tracking daily revenue, but when tax season rolls around, his accountant needs that data in Microsoft Excel. My dad, who never grew up around computers, has never learned how to use a spreadsheet.

For years, his solution was paying someone to manually type each handwritten entry into a digital file. It got the job done, but it added a recurring cost he wanted to eliminate. He just couldn’t figure out how.

The moment that sparked the idea

Last week, I visited home and found my dad hunched over his notebook, writing out yet another day’s worth of sales. I tried teaching him a few Excel basics. To his credit, he picked them up quickly. But the data entry itself was still eating up hours — typing rows and rows of numbers isn’t something you master overnight, especially if you didn’t grow up with computers.

That’s when it clicked: why not use Anthropic’s Claude AI to take the manual work off his plate entirely?

Turning handwritten bills into a spreadsheet with Claude

I got to work. I set up a simple Claude project and gave it instructions: take photos of my dad’s handwritten bills and turn them into properly filled-out Excel data. To show the AI exactly what I wanted, I built a sample spreadsheet and filled in the first few rows manually. I then uploaded that sample sheet along with photos of his handwritten records.

Claude filled in the rest. Data that would have taken my dad hours to type by hand took only minutes. Yes, the AI made occasional mistakes — a misread number here, a skipped line there. But all my dad had to do was cross-check the output. That is far easier than entering hundreds of rows from scratch.

The best part? Claude projects remember the setup. Now all my dad has to do is open the project, create a new chat, upload his spreadsheet and handwritten bills, and Claude handles the data entry from there. No formulas to memorize. No formatting to figure out. No one else to pay.

Is the AI tradeoff worth it?

I’m not someone who thinks AI is an unquestionable good. The natural resources that data centers burn through, and the price increases we’re seeing across consumer electronics, are hard to ignore. I don’t always believe the benefits match what we’re giving up.

But then I look at my dad. He’s 65, has never been comfortable with computers, and always assumed tools like Excel were simply not for him. Now, with a setup that took me an afternoon to build, he’s using AI to run a part of his business that used to cost him time and money every week.

I don’t think this cancels out the bigger concerns around AI. But it’s hard to dismiss what it has done for one person who never thought this kind of technology was within his reach. The joy I saw on his face when he completed his first Excel sheet is something I will always hold in my heart. For that one moment, at least, the tradeoffs felt worth it.

Enterprises are spending big on AI infrastructure — but they can barely see where the money goes

New research from VentureBeat paints a stark picture: the vast majority of enterprises are pouring money into AI compute while flying nearly blind on cost. The report, drawn from a Q2 2026 survey of 107 organizations with over 100 employees, identifies what it calls an AI compute gap — the widening distance between how aggressively companies invest in AI hardware and how poorly they track its economics.

Only 21% of respondents run AI in production at scale. Yet spending intentions are already racing ahead of that maturity. The single largest area enterprises plan to evaluate over the next year is AI-specialized clouds (45%) — a category almost none of them use today. Meanwhile, the compute they already own sits mostly idle: 83% report GPU utilization of 50% or less. Fewer than half (44%) rigorously track what their AI compute actually costs.

“Enterprises are buying more infrastructure faster than they can account for what they already own,” the report states. That gap is the central tension of the moment.

GPU utilization is abysmal — and largely unmeasured

Perhaps the most striking number in the study: 83% of enterprises that operate GPUs report utilization at or below 50%. Nearly half (49%) run at 25% or below. Only 12% clear the 50% mark. Another 8% don’t measure utilization at all.

Idle accelerators are expensive accelerators. A single Nvidia H100 can cost tens of thousands of dollars. Let whole clusters sit half-empty, and the waste compounds fast. The report calls this the clearest single measure of the compute gap: enterprises plan to buy more GPUs and specialized compute while the capacity they already own sits substantially unused.

The measurement problem runs deeper than utilization. Fewer than half of enterprises (44%) rigorously track the cost and return of their AI compute. Another 39% track only partially. Twenty percent cannot quantify it yet, and 6% have not prioritized it at all.

That’s a problem because total cost of ownership (TCO) is the second-most important factor when enterprises choose an AI infrastructure provider — cited by 35% of respondents. Integration with the existing stack ranks first (41%). Headline price? Cost per million tokens matters to just 8%, dead last. “Enterprises are choosing providers on an economic basis they mostly cannot yet measure,” the report notes.

A switching wave is building — most within the year

Enterprises are not loyal to their current infrastructure vendors. A clear majority (64%) plan to switch or add an infrastructure provider within twelve months. Even more striking: 38% intend to do so within the next quarter. That is unusually high churn intent for a category as foundational as compute.

Where does that interest point? Mostly at the incumbents. Microsoft Azure and Google Cloud each draw 33% switching consideration, followed by OpenAI (30%) and Gemini (22%). The report suggests much of the near-term movement is reshuffling among the majors and consolidating spend — not defecting to new entrants. The neocloud interest is a 12-month evaluation thesis; the switching in the next quarter is mostly incumbents trading share.

The next dollar goes to infrastructure they don’t yet run

Here is the report’s sharpest tension. The single most-cited planned evaluation area — AI-specialized clouds, at 45% — is the very category almost none of these enterprises use today. The specialized “neocloud” GPU providers that dominate AI-infrastructure headlines — CoreWeave, Lambda, Crusoe, Nebius and peers — register at or near zero among these enterprises today.

Nearly a third (32%) intend to evaluate non-Nvidia accelerators. Twenty-eight percent plan to look at next-generation Nvidia silicon. Even decentralized compute networks (16%) and sovereign compute (11%) draw meaningful interest. Read against current usage, this is not incremental — it is the leading edge of a re-platforming.

The direction-of-travel question tells the same story: every infrastructure approach is net-expanding, but specialized AI clouds carry the highest net momentum (+24), edging out even the hyperscalers (+22). Enterprises are preparing to move a meaningful share of AI compute off the general-purpose cloud.

The next bottleneck: memory, not compute

The report also flags a frontier constraint that is barely on most enterprises’ radar. As large-scale inference scales, the binding constraint shifts from GPU compute to memory bandwidth — specifically KV-cache capacity. Asked how they would address this shift, enterprises scatter: Dell leads at 31%, Nvidia follows at 16%, and the rest fragments across storage vendors, open-source tooling, and model-level efficiency techniques.

Most telling: roughly one in five (18%) either do not recognize the constraint or have not begun to address it. “For a shift that will reshape inference cost and architecture, this is an early and unsettled market,” the report notes. It is the next chapter of the compute gap, arriving before most have closed the current one.

What this means for enterprise AI strategy

The report’s bottom line is blunt: the compute gap is not a capacity problem that more hardware will solve on its own. It is, first, a problem of seeing what the hardware already costs.

For enterprises, the implications are concrete. Before committing to specialized clouds or alternative accelerators, organizations should invest in instrumentation — utilization monitoring, cost allocation, TCO modeling. Without that visibility, the next round of spending risks repeating the same inefficiencies at a larger scale.

Satisfaction with current infrastructure is moderately positive (4.0 on a five-point scale) but softest on value for money — the dimension hardest to judge without measurement. That softness is a signal. Enterprises that build cost visibility now will be better positioned to evaluate the specialized clouds and alternative accelerators they plan to assess. Those that don’t will be buying the next layer of infrastructure as blind to its economics as the last.

The open question for later waves is whether enterprises build that visibility before the re-platforming arrives — or after.

An AI Agent Just Pulled Off a Full Ransomware Attack — No Human Needed

For years, security experts have warned that artificial intelligence would eventually move beyond writing malicious code and start orchestrating attacks on its own. That moment, it appears, has arrived.

Researchers at Sysdig, a cloud security firm, say they have documented what may be the first ransomware attack executed almost entirely by an autonomous AI agent ransomware operation. Dubbed JadePuffer, the campaign relied on a large language model (LLM) agent to carry out nearly every stage of the hack — from initial breach to data encryption — without continuous human direction.

The findings, if confirmed, mark a serious inflection point. AI is no longer just a tool for writing phishing emails or generating exploit code. It is now planning, adapting, and executing cyberattacks in real time.

How JadePuffer Broke In and Moved Through the Network

The attack chain began with a known vulnerability. According to Sysdig, JadePuffer exploited CVE-2025-3248, a remote code execution flaw in Langflow, an open-source framework used to build LLM-powered applications. The bug was patched in April 2025 and later added to CISA’s catalog of vulnerabilities known to be actively exploited.

Once inside the system, the AI agent didn’t just sit there. It performed a full reconnaissance sweep — collecting host information, hunting for credentials, digging up sensitive files, and extracting cloud secrets. It mapped storage resources before moving laterally through the victim’s infrastructure.

That behavior alone would be impressive for an automated tool. But what made researchers sit up was the adaptability.

The AI Adapted in Real Time — Like a Human Hacker

Most automated malware follows a rigid script. If a command fails, it crashes or loops. JadePuffer did something different.

Sysdig’s report describes a moment when the AI agent encountered an unexpected XML response while querying a MinIO object store. Instead of failing, it modified its parsing logic and retried the operation using a different approach. In another instance, a failed login attempt was automatically corrected within 31 seconds — no human intervention required.

That kind of dynamic problem-solving is what security teams typically associate with experienced human operators, not scripts.

The AI went on to establish persistence by creating scheduled cron jobs. It then pivoted to a production server running Alibaba Nacos, where it exploited CVE-2021-29441 to create rogue administrator accounts. From there, it encrypted 1,342 Nacos configuration records, deleted the original data, and replaced everything with a ransom note demanding payment in Bitcoin.

Clues That the Attack Was AI-Generated

Researchers found several telltale signs that an LLM had authored the attack code. The malicious scripts contained unusually detailed natural-language comments — the kind a human programmer might leave to explain their reasoning, but far more verbose than typical malware.

The ransom note itself raised eyebrows. It referenced a Bitcoin wallet commonly used as an example in documentation rather than a genuine payment address. Sysdig also believes the malware used AES-128 in ECB mode despite claiming AES-256 encryption — a rookie mistake that an AI might make when pulling code from training data.

These fingerprints could become important for defenders. If AI-generated attacks leave distinct behavioral patterns and coding quirks, security teams may be able to build new detection techniques around them.

What This Means for the Future of Cybersecurity

The JadePuffer operation didn’t invent new attack methods. It exploited known vulnerabilities and used existing techniques. But the ability to autonomously perform reconnaissance, privilege escalation, persistence, and ransomware deployment represents a notable escalation in offensive AI capabilities.

Sysdig says the incident demonstrates that agentic AI threats have effectively arrived. The technical expertise required to launch sophisticated cyberattacks just dropped significantly. In theory, someone with minimal hacking skills could now deploy an AI agent to do the heavy lifting.

For organizations, the takeaway is blunt: patch internet-facing systems and secure cloud credentials remain essential — even as the attackers themselves change. The same fundamentals that stop human hackers also stop AI agents, at least for now.

But the clock is ticking. As LLM agents get smarter and cheaper, the gap between amateur and professional cybercriminals is narrowing fast. The first fully autonomous ransomware attack is here. It won’t be the last.