The day AI became the reason finance cut jobs── Job cuts and a huge fundraise, side by side on the same day
Download the video (MP4, 10 MB)
Some companies now name AI as the reason they are letting people go. On the same day, the firms that build AI were said to be drawing enormous sums. For working people, that is unsettling news. But I want to ask a single question. Of the big stories we hear, how much has actually happened, and how much is only forecast?
Drug makers, too, now decide more often when to bring in AI. If you judge by the size of a headline, you will get it wrong. Headcounts and money both mix solid facts with softer ones. Today, let us follow the day's events and keep those two apart.
01Job cuts at finance majors
Source 매일경제 ／ Demokratis.co.id ／ Business Insider ／ Wocult
We start with the industry that handles money. Why do staff reductions stand out there?
| Company | Scale | Reported reason |
|---|---|---|
| American Express | 5% of employees | To prepare for AI |
| Barclays | 5,000 staff | Shifting spending to AI |
| A payments major (not named in headline) | Not in headline | Cuts attributed to AI |
| Uber | Layoffs last month | Then expanded ChatGPT for employees |
One tally counts over 30 firms tying job cuts to AI (Wocult)
Most of these firms process huge volumes of paperwork and calculation. Headcount is starting to move where machines can most easily take over tasks. But naming AI as the reason is not the same as AI doing the work. Some firms may use the word to explain costs or a slow economy. Still, it pays to tell the two apart. I care less about how many left than about what those who stay are given.
Follow this story step by step. Money once paid to people is being moved to pay for machines instead. Which roles suffer most, though, is still a guess. Firm facts and guesses sit side by side in a single account.
Pharma also spends many hands on writing documents and checking them. If the same shift comes, the first question is who checks what the machine produces. Keep reviewers in place before you reduce staff. And record what was checked and who decided, so you can explain it later.
So how far, and how large, could this spread? Next, the people who tried to put a figure on it.
02Job estimates and Congress
Source Startup Fortune ／ Fortune ／ The New York Times
Wanting to know the future scale is natural. But every estimate rests on assumptions. Change them, and the answer changes too.
Goldman Sachs
Estimated the share of US workers that AI could eventually displace. This site's morning edition reported it as about 10%.
Monthly job losses
Fortune reported analysts saying AI is eliminating over 10,000 jobs a month and that 9% of jobs are at risk.
US Congress weighs options
Fearing mass AI layoffs, some members of Congress are reportedly looking at ways to support workers.
These three items point the same way. Yet how solid they are differs widely. One is a ceiling on what might happen someday. One counts work being lost right now. The last is politics still under discussion. Heard together, they sound like a settled future. In fact, they are claims of very different certainty.
Treat a future ceiling and a present count as equal, and you will misjudge. An estimate can swing widely when its assumptions shift. Even so, I give more weight to lawmakers starting to act. Whatever the figures turn out to be, society has begun to prepare.
This is not a distant story for people in pharma HR and training. Which tasks go to machines, and which stay with people? Draw that line yourselves, or outside figures will push you around. Outside estimates do not know your work. Review your own tasks one by one, and test what can be handed over.
So how cheap, and how reliable, has the technology that takes over this work become?
03A model that needs less memory
Source Tom's Hardware ／ Bloomberg.com ／ Reuters
For machines to do people's work, running them must get cheaper. On that cost, news came from a developer in China.
| Player | Move | Details |
|---|---|---|
| DeepSeek | New model | 1/4 the HBM and 1/8 the SSD storage |
| DeepSeek | Remaining issue | Hallucinations in agentic workloads |
| Moonshot | AI upgrade | First since its own "DeepSeek moment" |
| DeepSeek, MiniMax | Use abroad | State media-affiliated entity approved global use, per report |
One developer's new release needs far fewer of the costly parts that run it. The same work can move to much cheaper machines. That should let more companies adopt it. Smaller firms priced out until now may get a chance. A report of approval for use outside the country also suggests a push abroad, selling on low price.
Less reliance on costly parts may shift who sets the price of this technology. The makers of those parts could slowly lose leverage. That is my own view, and it is not yet confirmed.
Yet confident errors have not gone away. Cheaper does not mean more correct. Cost and accuracy must be measured separately.
In pharma documents, one error can touch patient safety. The cheaper the tool, the more human checking matters, in my view. Put part of the savings into review. Choose on price alone, and fixing mistakes later may cost more.
While some developers compete on price, others gather vast funds and aim to win on scale.
04Reports on OpenAI's fundraising
Source TechCrunch ／ Bloomberg.com
Talk of huge funding moves the mood of markets. So let us look at how outlets described the same story.
- Reported
- $30B new raise at a $1.25T valuation
- Format
- Article
- Reported
- OpenAI trimmed its target valuation
- Format
- Headline of a video program
This site's evening edition noted that the reports conflict
Two stories on one company's funding seem to point opposite ways. One describes raising money at a high price. The other says the goal was lowered. Which is right, we cannot yet say. We only know the deal is not settled. When reports split, wait for the terms instead of picking a side.
The big figure in the headline is not cash flowing into the company. It shows what investors and the company think the whole firm is worth. Bigger hopes mean a bigger figure.
Before shares reach the market, this estimate is especially unstable. Once listed, the price is set daily, in plain view. Until then, it is a private agreement among a few.
When drug makers pick a partner here, its financial depth matters. But a reported sum alone is no reason to relax. If its funding wobbles, the cost of your tools and how long they are offered may change. Wait for fixed terms before you decide.
The same way of reading applies to the sums in a drug maker's own deals. Let us look inside one next.
05Lilly's drug delivery deal
Source Digital Health News
Now to a company that makes medicines. How much of the reported sum is actually promised?
| Item | Amount or content | Nature of the sum |
|---|---|---|
| Partner | METiS (Hong Kong) | Partnership to find AI drug delivery technology |
| Upfront | Over $10M | Paid first |
| Maximum | $1.1B | Ceiling, not a fixed sum |
Most of the headline sum is milestones, paid at each research stage
A major American drug maker has teamed up with a young tech firm. The goal is to use machine learning to send drugs to the right place in the body. The large sum in the reports is not money set to be paid. It is the ceiling if the work goes well.
Most such ceilings are payments due once the work hits agreed checkpoints. Miss a checkpoint, and that money never moves.
Which particles will carry a drug to its target? I think the plan is to narrow the candidates with machine learning. But I read that from the headlines. I have not confirmed it in the deal terms.
When you read partnership news, look at the first payment, not the ceiling. That shows what the partner truly committed today. Note one more thing. Machine learning is entering not only drug discovery, but also how drugs reach their target.
Meanwhile, a government moved to limit this kind of cross-border investment itself. How research gets funded may change.
06On the pharma front line
Source Fortune ／ Loeb & Loeb LLP
National rules now join the flow of money. Of today's stories, this one sits closest to pharma work.
The proposal groups two fields together, computing and the science of living things. It then watches money headed to certain countries, even small amounts. Drug research and machine learning are starting to be treated alike, as technology to protect. But this is still a draft, not a decision.
| Body | Content | Stage |
|---|---|---|
| US Treasury | AI and biotech investment in China, Russia | Rule proposed |
| Pennsylvania | Law requiring AI disclosure | Enacted; scope in question |
Also that day, one state passed a law requiring people to reveal their use of the technology. Yet what it covers is still unclear. More rules mean more work to interpret them. Users are left asking what to reveal, and how much.
If it is finalized, funding for overseas labs and joint research could face review. Depending on the country and field, familiar ways of partnering may no longer be allowed.
Pharma often builds research across borders. Before the draft is final, map which of your partnerships and stakes it might touch. Legal and research teams will need to sit down together more often.
As protection widens, a quieter story was unfolding about ways to check how these systems work.
07What is being overlooked
Source Anthropic ／ The Guardian ／ 24/7 Wall St.
Quiet news often matters later. To finish, let us pick up stories that are easy to miss.
Formal verification post
Anthropic said on its official blog that LLMs are making large-scale formal verification feasible.
Doubts on internal report
Experts reportedly questioned Anthropic's report on how AI has changed work inside the company.
AI agents trading
Coinbase now lets AI agents trade in crypto accounts. Risks have been pointed out by observers.
These three share one question. Who checks the machine's work, and how? One is about proving software correct with mathematics, made practical by language models. One is outside experts challenging a company's account of how its own work changed. The last is a service that lets software buy and sell assets on a person's behalf.
If correctness can be proven mathematically, people need not read every output by hand. But this claim comes from the developer itself. Outside checks are still to come. The checking method is itself still being tested.
Pharma work rests on keeping records of what was checked. As machines take on more, how you build those records will be questioned. The doubted report is a lesson in how to show your own results to outsiders. Automated trading is not someone else's problem either. When software acts for people, you need a way to stop it and a trail to follow.
All of today's events led to one point. Read checked facts and records apart from forecasts.
Will OpenAI's funding terms be settled, and which way will the conflicting TechCrunch and Bloomberg valuation reports settle?
Open the full transcript
Intro
Some companies now name AI as the reason they are letting people go. On the same day, the firms that build AI were said to be drawing enormous sums. For working people, that is unsettling news. But I want to ask a single question. Of the big stories we hear, how much has actually happened, and how much is only forecast?
Drug makers, too, now decide more often when to bring in AI. If you judge by the size of a headline, you will get it wrong. Headcounts and money both mix solid facts with softer ones. Today, let us follow the day's events and keep those two apart.
CH 01 Job cuts at finance majors
We start with the industry that handles money. Why do staff reductions stand out there?Most of these firms process huge volumes of paperwork and calculation. Headcount is starting to move where machines can most easily take over tasks. But naming AI as the reason is not the same as AI doing the work. Some firms may use the word to explain costs or a slow economy. Still, it pays to tell the two apart. I care less about how many left than about what those who stay are given.Follow this story step by step. Money once paid to people is being moved to pay for machines instead. Which roles suffer most, though, is still a guess. Firm facts and guesses sit side by side in a single account.Pharma also spends many hands on writing documents and checking them. If the same shift comes, the first question is who checks what the machine produces. Keep reviewers in place before you reduce staff. And record what was checked and who decided, so you can explain it later. So how far, and how large, could this spread?
Next, the people who tried to put a figure on it.
CH 02 Job estimates and Congress
Wanting to know the future scale is natural. But every estimate rests on assumptions. Change them, and the answer changes too.These three items point the same way. Yet how solid they are differs widely. One is a ceiling on what might happen someday. One counts work being lost right now. The last is politics still under discussion. Heard together, they sound like a settled future. In fact, they are claims of very different certainty.Treat a future ceiling and a present count as equal, and you will misjudge. An estimate can swing widely when its assumptions shift. Even so, I give more weight to lawmakers starting to act. Whatever the figures turn out to be, society has begun to prepare.This is not a distant story for people in pharma HR and training. Which tasks go to machines, and which stay with people?
Draw that line yourselves, or outside figures will push you around. Outside estimates do not know your work. Review your own tasks one by one, and test what can be handed over. So how cheap, and how reliable, has the technology that takes over this work become?
CH 03 A model that needs less memory
For machines to do people's work, running them must get cheaper. On that cost, news came from a developer in China.One developer's new release needs far fewer of the costly parts that run it. The same work can move to much cheaper machines. That should let more companies adopt it. Smaller firms priced out until now may get a chance. A report of approval for use outside the country also suggests a push abroad, selling on low price.Less reliance on costly parts may shift who sets the price of this technology. The makers of those parts could slowly lose leverage. That is my own view, and it is not yet confirmed.Yet confident errors have not gone away. Cheaper does not mean more correct. Cost and accuracy must be measured separately.In pharma documents, one error can touch patient safety. The cheaper the tool, the more human checking matters, in my view. Put part of the savings into review. Choose on price alone, and fixing mistakes later may cost more. While some developers compete on price, others gather vast funds and aim to win on scale.
CH 04 Reports on OpenAI's fundraising
Talk of huge funding moves the mood of markets. So let us look at how outlets described the same story.Two stories on one company's funding seem to point opposite ways. One describes raising money at a high price. The other says the goal was lowered. Which is right, we cannot yet say. We only know the deal is not settled. When reports split, wait for the terms instead of picking a side.The big figure in the headline is not cash flowing into the company. It shows what investors and the company think the whole firm is worth. Bigger hopes mean a bigger figure.Before shares reach the market, this estimate is especially unstable. Once listed, the price is set daily, in plain view. Until then, it is a private agreement among a few.When drug makers pick a partner here, its financial depth matters. But a reported sum alone is no reason to relax. If its funding wobbles, the cost of your tools and how long they are offered may change. Wait for fixed terms before you decide. The same way of reading applies to the sums in a drug maker's own deals. Let us look inside one next.
CH 05 Lilly's drug delivery deal
Now to a company that makes medicines. How much of the reported sum is actually promised?A major American drug maker has teamed up with a young tech firm. The goal is to use machine learning to send drugs to the right place in the body. The large sum in the reports is not money set to be paid. It is the ceiling if the work goes well.Most such ceilings are payments due once the work hits agreed checkpoints. Miss a checkpoint, and that money never moves.Which particles will carry a drug to its target?
I think the plan is to narrow the candidates with machine learning. But I read that from the headlines. I have not confirmed it in the deal terms.When you read partnership news, look at the first payment, not the ceiling. That shows what the partner truly committed today. Note one more thing. Machine learning is entering not only drug discovery, but also how drugs reach their target. Meanwhile, a government moved to limit this kind of cross-border investment itself. How research gets funded may change.
CH 06 On the pharma front line
National rules now join the flow of money. Of today's stories, this one sits closest to pharma work.The proposal groups two fields together, computing and the science of living things. It then watches money headed to certain countries, even small amounts. Drug research and machine learning are starting to be treated alike, as technology to protect. But this is still a draft, not a decision.Also that day, one state passed a law requiring people to reveal their use of the technology. Yet what it covers is still unclear. More rules mean more work to interpret them. Users are left asking what to reveal, and how much.If it is finalized, funding for overseas labs and joint research could face review. Depending on the country and field, familiar ways of partnering may no longer be allowed.Pharma often builds research across borders. Before the draft is final, map which of your partnerships and stakes it might touch. Legal and research teams will need to sit down together more often. As protection widens, a quieter story was unfolding about ways to check how these systems work.
CH 07 What is being overlooked
Quiet news often matters later. To finish, let us pick up stories that are easy to miss.These three share one question. Who checks the machine's work, and how?One is about proving software correct with mathematics, made practical by language models. One is outside experts challenging a company's account of how its own work changed. The last is a service that lets software buy and sell assets on a person's behalf.If correctness can be proven mathematically, people need not read every output by hand. But this claim comes from the developer itself. Outside checks are still to come. The checking method is itself still being tested.Pharma work rests on keeping records of what was checked. As machines take on more, how you build those records will be questioned. The doubted report is a lesson in how to show your own results to outsiders. Automated trading is not someone else's problem either. When software acts for people, you need a way to stop it and a trail to follow. All of today's events led to one point. Read checked facts and records apart from forecasts.
Wrap-up
The louder a story about lost jobs or new money, the more it needs checking. Separate what happened from what is forecast. Then you can see what pharma should prepare for. Tomorrow, I will watch which way the split funding stories land. That will help us judge how far to trust the sums we read about.
- CH 01매일경제「英国系投資銀行バークレイズが今年、職員5000人を減員する。 銀行が全事業部門にわたって人工知能(AI)を導入し、開発者が減員の大きな打撃を受けるものと見られる。22日（現地時間）、英フィナンシャル」 mk.co.kr
- CH 01Demokratis.co.id「American Express Cuts 5 Percent of Workforce to Prepare for AI」 demokratis.co.id
- CH 01Business Insider「Uber doubles down on ChatGPT for workers after last month's layoffs」 businessinsider.com
- CH 01Wocult「Over 30 firms have now tied job cuts to AI, with HSBC and Amazon leading」 wocult.com
- CH 02Startup Fortune「Goldman Sachs says AI could eventually displace one in eleven US workers」 startupfortune.com
- CH 02Fortune「AI is killing 10,000-plus jobs per month and 9% of jobs are at risk, analysts say」 fortune.com
- CH 02The New York Times「Fearing Mass A.I. Layoffs, Some in Congress Eye Ways to Aid Workers」 nytimes.com
- CH 03Tom's Hardware「Deepseek's latest AI model is smarter than ever, but its low running costs come at a price — uses only one-quarter of the HBM and one-eighth of SSD storage but still battles hallucinations in agentic workloads」 tomshardware.com
- CH 03Bloomberg.com「Moonshot Debuts First AI Upgrade Since Its Own DeepSeek Moment」 bloomberg.com
- CH 03Reuters「China's DeepSeek, MiniMax win approval for AI global use, state media-affiliated entity says」 reuters.com
- CH 04TechCrunch「OpenAI working on new $30B raise at $1.25T valuation ahead of IPO」 techcrunch.com
- CH 04Bloomberg.com「Watch OpenAI Trims Target Value & Volvo Cars Cuts 2,500 Jobs」 bloomberg.com
- CH 05Digital Health News「Eli Lilly Partners with METiS TechBio in Up to $1.1 Bn AI Drug Delivery Deal」 digitalhealthnews.com
- CH 06Fortune「Treasury proposes rule to crack down on AI, biotech investments over $50,000 in China, Russia」 fortune.com
- CH 06Loeb & Loeb LLP「A New AI Disclosure Law from Pennsylvania. But What Does It Actually Cover?」 loeb.com
- CH 07Anthropic「LLMs are making formal software verification feasible at a much larger scale」 anthropic.com
- CH 07The Guardian「Experts cast doubt on Anthropic report detailing how AI has altered work at the company」 theguardian.com
- CH 0724/7 Wall St.「Coinbase Now Lets AI Agents Trade in Your Crypto Account. What Could Go Wrong?」 247wallst.com
Articles used
- AI Daily News 2026-10-24
- AI Daily News 2026-10-23 Evening
- AI and the Pharmaceutical Industry 2026-10-24
Today's related reports
- AI Daily News October 24, 2026
- AI & Economy News October 24, 2026
- AI & Finance News October 24, 2026
The narration is synthetic speech. The content draws only on this site's articles from the same day. Each edition passes seven plain-language checks before publication. That means known obstacles to comprehension are held below threshold — it is not a guarantee of comprehension.