Category Archives: rants

There is NO Infinite Compression – The Latest DeepSeek Paper is BullCr@p!

Every decade or so, some idiots who never studied Huffman coding or Information Theory believe they have cracked the problem of infinite compression, and this linked paper is just the latest example of this lunacy. I really hope this was a joke paper authored by AI because it’s all bullcr@p!

On average, a text token in a LLM should require 20 bits or less (as 17 bits support a 129,000 word vocabulary) while a vision token can be 16,384 bits (based on 1024 dimensional continuous vectors) — because it takes a lot of bits to represent pixelation of a square in a 2-D image! This says you can store about 820 text tokens in the same space it takes to store one vision token. Or, you can store the entire text (lossless) in 48K, versus the 4M it would take to store the 250 vision tokens (using very lossy compression) that are required in the paper. Looks like a LOT of people can’t do basic math if this is being praised as revolutionary!

Moreover, the raw text, which maintains the full context if the tokens are kept in order, is not only fully lossless, but can be compressed using a modified Lempel-Ziv algorithm to take up an average of less than 2 bits per character (and achieve up to an 80% compression rate). Given that the average length of a word in average text is 5 characters, and a space is one character, 2500 words would be 15,000 characters, storable in 30,000 bits or a mere 4K! In other words, this paper is trying to pass off a ONE THOUSAND FOLD increase in space requirements as space saving! Pure lunacy!

In other words, if someone is claiming something too good to be true, it is! Don’t fall for it or the sure to follow claims that DeepSeek OCR is revolutionary because of this. (Since every document is different, you can’t imagine the true loss with a 90% vision token reduction!)

Dangerous Procurement Predictions Part II

As per our first post, if you read my predictions post, you know SI hates predictions posts. It fully despises them because the vast majority of these posts are pure optimistic fantasy and help no one. Why are the posts like this? Because no one wants to hear the sobering reality off of the bat in the new year and the influencers care more about clicks than actually helping you.

But the predictions are not only bad, they’re dangerous. And to make sure you don’t fall for them and make bad decision based on them, we’re going to tackle some of the most dangerous predictions, which include predictions that look innocuous at first glance (like the last prediction on how a big legacy suite will go out of business) but hide the dangerous consequences of what will actually happen if a big suite finds itself in big trouble. Today we tackle the next four, and you can be sure this won’t be the last post in our series. Feeds are still being flooded with prediction posts, and I’m done ignoring the insanity.

4. The jobs market will be tough for the first half of the year, but will start to pick up in Q3 and Q4.

The job market is tied to the economy, and everyone predicts the job market will rebound when the economy picks up. But here’s the thing. Even when the economy picks back up, the job market never does quite as well as the last time. And the economy isn’t going to magically improve half-way through the year. This is the exact same thing we’ve been told the last two years, and it hasn’t happened.

First off, most of the first world economies around the world are flat, borderline recession, or in recession. Secondly, the only thing propping the US economy up right now is AI, and the money circles keeping it afloat as all the AI, Hardware, and Software companies keep moving the same money around investing in each other to keep each other afloat. If the bubble bursts, the US is in trouble, and the economy will quickly flush itself down the toilet. And the job market will go with it.

Considering only the big tech giants who have been hoarding cash for the last few years are in good shape, and everyone else is trying to conserve cash to survive not only the current market but a potential recession, the last thing they are going to do is hire unless absolutely necessary to fill a critical role as a result of a departure. Remember, they’ve spent the last two years using AI as an excuse to lay people off and are always looking for the next excuse to lay people off, not hire them!

Jobs will continue to be super scarce, and only the best will have a chance to land one.

5. We’re in the early stages of a broader pushback (against unnecessary upgrades or technology investments).

A few companies smartening up and saying no to forced big provider upgrades, eight (8) figure consultancy projects, and big Gen-AI investments is not pushback. There have always been a few leaders who have broken away from the pack, did the math, and made the right decisions, but the pack is still charging ahead on Gen-AI. Every big software shop except IBM (who hired a CEO who can actually do math) has invested heavily in Gen-AI, which still loses four dollars for every dollar of revenue, despite any hopes of a real return in the near future and a 94% failure rate.

Let’s face reality. I warned this space about The Vendor In Black nineteen years ago and how he always Comes Back sixteen years ago, no one took heed then, and no one is taking heed now. The business model of the enterprise software space, which has not changed for the two decades I’ve been covering it, is to solve the problem created by the old sh!t by selling the customers the new sh!t that comes with new problems so they can sell even newer sh!t in three years to fix those (and so on). Same old story. Only the vendor names change.

6. We Won’t Buy Things; We’ll Orchestrate Ecosystems.

This prediction likely came straight from the A.S.S.H.O.L.E. and anyone who repeats it should be ashamed of themselves. There are no AI Employees. Claims to the contrary are false and anyone making those demeaning and degrading claims is simply dehumanizing you. And, as we have clearly explained, you definitely don’t want agentic buying because it will happily spend your money not only on stuff you don’t need but stuff that doesn’t exist and, if you’re super unlikely, stuff that is highly illegal. You need wood, it will buy up all the Minecraft wood because it’s cheap and call your problem solved. And that’s if you’re lucky. If you’re not, it will fulfill your resin need with an illegal purchase of hash (the drug) on the dark web (which is labelled resin so the poster can claim they never advertised an illegal drug). And so on.

Plus, as we have already noted, most of today’s “orchestration” platforms in Source-to-Pay are really ORCestration platforms and can barely connect a handful of major Source-to-Pay offerings. They’re nothing close to what is needed to orchestrate ecosystems.

7. Boards will Zero in on Supply Chain Security and Supplier Risk shifts from quarterly PowerPoints to continuous “signalops”.

Just like they won’t invest more in cybersecurity, they won’t invest more in supply chain security until they lose a shipment in the tens of millions. After all, they’ve got supply chain insurance, why should they care? Especially since their current security measures have been sufficient up until now.

But here’s the thing. When the economy goes down, jobs go down. And then two things happen. People get desperate and turn to crime. And criminals, when their investments in drugs, alcohol, gambling, prostitution, and other quasi-legal through illegal activities start losing money because unemployed people run out of money to spend on their vices, these criminals get desperate too — and high value theft becomes more attractive. A temporarily unguarded truck here. A container there. An entire warehouse. And so on.

If it’s critical raw materials they can move (like rare earths), in-demand finished electronics they can sell (like iPhones, where a single container will contain at least 20M worth), military equipment or weapon (component)s that are now in demand globally, they’ll take bigger and bigger chances, especially if there are weaknesses in security. It’s not just cyber attacks that are going to increase, it’s physical attacks, supply chains aren’t ready, and companies won’t even stop preparing them until they lose tens of millions, don’t recover it all through insurance, and risk losing their insurance entirely. No one likes the math of risk prevention because, when it works, you don’t see the return. Even though it’s so much cheaper than insurance! And that’s why, in the majority of organizations, nothing will change.

Dangerous Procurement Predictions Part I

If you read my predictions post, you know SI hates predictions posts. It fully despises them because the vast majority of these posts are pure optimistic fantasy and help no one. Why are the posts like this? Because no one wants to hear the sobering reality off of the bat in the new year and the influencers care more about clicks than actually helping you.

But given how dangerous and costly the hopeful fantasy has become, not only did SI swallow its disgust and give you a realistic predictions post, but it’s going to collect and lay bare the most dangerous of the predictions that, even if seemingly innocuous, will lead you astray if you believe them. And now some of the influencers and LinkedIn aficionados are taking up the claims, and the charge, but like many other claims, they are overstated.

Today we tackle the first three, but you can expect this to be the first of many posts as dangerous prediction posts flood your feeds for the rest of the month.

1. The “Great Convergence” Accelerates

The claims of of the ORChestration providers is that all roads lead to them, the convergence will accelerate, and you won’t have to worry about what you need because, as long as you have orchestration, you’ll have it all!

For example, if you want to use the largest orchestration provider in S2P, your are limited to the platforms they have already integrated. The same goes for the second or third largest. Plus, if the providers you want to integrate aren’t reasonably sized Source to Pay providers, good luck expecting the workflow to support them appropriately.

Moreover, they were built to minimally support the existing solutions, not emerging solutions in the Source to Pay and extended Supply Chain Marketplace. In other words, the convergence will continue at a snails pace, but it will never be great!

2. “X” Finally Gets Modern Attention

It doesn’t matter what X is — if X has been needed, but ignored, for the last ten years, it’s NOT going to all of a sudden be addressed this year. For whatever reason, it will continue to be ignored.

Example #1, Cybersecurity.

As per my recent post on breaking down the risks: IP / cyberattacks, the risk of cyberattacks has been high since 2014, a year when 71% of organizations were affected by a successful cyberattack! Ten years later, 70% of small to medium sized businesses are still getting hit by cyberattacks. (Which means that if it was going to get major attention, shouldn’t 2014 have been the year?!?)

Nothing has changed — the reason? Cybersecurity is seen as a cost, not a return. So, when a successful attack results in significant losses, organizations spend on improved cybersecurity, and ignore it until the next significant successful attack hits, and that is the only time they will spend for new systems across the board, and that’s it. That’s why cybersecurity, inside and outside the organization, won’t get any more attention this year than last year.

Example #2, Risk Management.

There’s a big reason it’s been the exact same risks in the state of procurement studies and reports for at least the last five, if not the last ten, years. It’s because, despite the fact that risks keep increasing, no one ever does anything about it … there’s no additional investment in risk management software. Why? Again, it’s seen as a cost and not an investment. And when you’re already paying for insurance, why pay for what, at best, seems like more?

Even though the cost of insurance will soon be unaffordable given that natural disaster and fraud losses are going through the roof, if you can even get insurance at all, risk management solutions are still being ignored by every organization that hasn’t suffered a major loss as a result of a risk-related event. (And who knows if insurance will cover AI losses when AI escapes the vending machine? It’s a question you should definitely be asking!)

Example #3, Direct.

That’s supply chain, right? Right?

Wrong! But that’s the view that the vast majority of Source-to-Pay providers have taken since the beginning. Sure a few big suites picked up a few smaller players that specialized in direct sourcing, but that’s about it from the big players. And there are a few startups here and there, but they’re all overlooked, underfunded, and not getting any traction.

Because it’s hard. Damn hard. And the majority of S2P players don’t want hard. They want easy. They built easy. They sell easy. And that’s all they want to do. (And, often, all they can do!)

We could continue, but you get the point.

3. One of the big legacy S2P suites will go out of business.

This is a prediction straight from the genius of Gary Wright. Only a Dream Weaver would predict this! This has happened exactly once since our space began in the late 1990s, and it wasn’t exactly going out of business, it was a big acquirer deciding the space wasn’t profitable enough and shutting the vendor down. Specifically, it was IBM shutting down Emptoris and shunting all the customers to SAP Ariba in 2017.

Every big provider in this space is controlled by PE who have poured tens, hundreds, or thousands of millions (that’s billions) into the firm. If it starts losing money, and if they think they can’t turn it around, rather than shutting it down, they’ll flip it to another firm at a loss (to recover some investment) who will pick up some fire sale acquisitions, integrate them, update the UX, install a whole new management team, fluff it up, rebrand it, and bring it out with a whole new spin. Like ERPs, Suites never die. Even if they’re twenty years behind the times.

So if a new big player hits the scene, check under the covers, do a bit of research, and dig up those skeletons. PE knows how to make everything old new again, but tech is not like fashion, and you don’t want two decades old SaaS, as that’s just the same old sh!t.

You Want Predictions? You Got Predictions!

Just Remember, You Asked For This!

If you’ve been following SI, you know SI hates prediction posts, because all they end up being is hopeful fantasy because no one wants to hear the sobering reality off the bat in the new year.

But given how dangerous and costly the hopeful fantasy has become, SI is going to swallow its disgust and give you what you want for — a realistic prediction post for 2026.

1) Gen-AI hype will continue for most, if not all, of the year. The market has figured out how to not only maintain the bubble, but blow it bigger than it’s ever been blown before. So despite all the “burst is imminent” predictions, the market, and the US in particular, is going to miraculously maintain, and even grow, the bubble. (As Mr. Stephen Klein has posted multiple times, they’ve got it all figured out!

2) On the flip side, the failure rate from (Gen-)AI-first solutions will remain above 90%. Almost-back-to-back Gen-AI studies in late 2025 by MIT and McKinsey found a 95% and 94% failure rate respectively. This will improve slightly, but based upon other indicators I’ve seen, stay around 92%.

3) I2O won’t save you — especially in direct!. Yes, Joël Collin-Demers, it is true that most of the big players I2O (Intake-to-Orchestrate) players will add “direct” support, but this “support” will basically be limited to BoM (bill-of-material) import and the ability to bulk buy the components on a BoM by BoM basis. And it will be less powerful than leaders like DirectWorks (now Ivalua), Pool4Tool (now Jaggaer), and EffiGo had a decade ago! Direct requires a lot more than bulk-buying a bill of materials. A LOT more!

4) Supply Chain risks (esp. around Wars, Tariffs, and Natural Disasters) will continue to accelerate, as well as losses due to (Cyber)Fraud, Natural Disasters, and Stock-outs, but investments in risk mitigation and management will continue to be minimal.

5) The same lack of investment goes for supplier/third party 360 solutions for proper integrated risk-and-compliance assessments up front during on-boarding. This is becoming more and more critical because even good suppliers can be risky when you dive into the factory locations, shipping lanes, ownership, etc.

6) Talent will continue to be a huge concern, but instead of actually investing in people, companies will continue to invest in “AI automation” in the hopes that they can fill the gap with tech (even though the tech has zero intelligence and is only as good as the real world experience of the techs who code it).

7) The M&A mania will hit a high about mid-year as the PE firms compete for the few remaining solutions with “unicorn” potential while simultaneously scooping up as many BoB assets to fill holes and acquire talent as the fire sales hit fast and furious. (Remember, I’m projecting double to triple the usual failure rate over the next year or so, 10% to 15% as compared to the typical 5%. We ended 2025 close to 6.5%, but after two years of depressed sales and all the (over)funding going to “AI” startups, a lot of good, niche, solutions are going unfunded and unnoticed, rapidly running out of funds, and need to get bought or shut their doors.

8) “Alt Suites” continue to proliferate as mid-market vendors and PE firms try to roll up their point-based vendors / BoBs (Best of Breeds) into a process/alt-function based offering that they feel will give the newly created vendor a unique angle to get attention (and sales).

9) Software-backed services continue to gain momentum as mid-market consultancies fight for limited budgets and large consultancies try to keep costs down, but in the interim, expect more Deloitte-style failures (where they’ve been caught twice publishing AI-generated cr@p as human analyst work — in Australia and Canada) among firms trying to cut costs with Gen AI-based software.

10) The Big Consultancies and Big Analyst Firms won’t help you! They’ll all publish their annual studies, which, as we pointed out in our You Don’t Need to Read Another State of Procurement Study for the Next 5 Years, but 90% of the “results” will be the same as this year, with just a re-ordering of the barriers, concerns, risks, etc. The only differences will be the “risk-du-jour” and “tech-du-jour”, which will likely be a new spin on supply chain risk and Agentic AI, but the conclusions will be “engage them to help you with a tech-du-jour plan that will solve all your problems” but in reality be a “let’s do a massive project to underpin your entire Procurement operation with tech-du-jour that will require a year of process definition of study, a year of implementation, and then years of training to customize the tech to your processes (i.e. train the tech enough so it actually works okay” that will cost you millions (upon millions) of services dollars but not actually deliver any ROI in the short (and even mid) term.

So, now that you finally got a realistic 2026 Procurement predictions post — tell me, how does it make you feel?

CEOs are hugely expensive. Why not automate them?

As per Will Dunn, as published on The New Statesman

Especially when hiring a CEO who doesn’t understand what makes the business profitable loses Billions:

Starbucks Loses 30 Billion

and doesn’t understand what is critical to the company product to the point costs can never be cut no matter how high those costs may look on the spreadsheet because the net result is not only product failure, but grounding/banning of your product and expensive lawsuits that costs Billions:

Boeing lost 11.8 Billion in 2024

After all, if we’re hiring CEOs without any relevant experience, actual business intelligence, or even logic, then why not use Artificial Idiocy? It’s not like the occasional hallucinations will be any worse that an average CEO’s these days (who believes investing Billions on empty promises is a good idea) … and the actual compute costs, even if in the six figures, will still be a tenth (or [much {much}] less) of what a CEO salary and benefit package actually costs!

So if you insist on creating fictional “AI Employees”, why not kick off 2026 by starting with a job that, sadly, Gen-AI agents can actually do?