TheChek News Flow

On Halloween, a Reddit user asked OpenAI CEO Sam Altman whether “GPT-5” was coming, as well as the full version of the company’s o1 reasoning model—a preview version of which went on sale seven weeks ago.

In other words, when would the public see the company’s next flagship large language model?

Altman answered, “we are prioritizing shipping o1 and its successors,” and added that limited computing resources meant that it was difficult to launch too many things in parallel.

We didn’t think much of it at the time. But now we have a better understanding of Altman’s comment—and why he is focused on launching an o-branded reasoning model rather than another version of GPT. That acronym stands for a generative pretrained transformer model (aka an LLM) that became the bedrock of ChatGPT and most other generative artificial intelligence products.

The reason? The pace of improvement of GPTs is slowing down, as we reported Saturday, with new details about the development of Orion, the next big GPT. And now that OpenAI has the ability to bake reasoning into GPTs such as Orion, it may start branding these models as part of the “o” family.

The prior flagship LLM, GPT-4, had publicly launched in early 2023 and sent shockwaves around the tech industry because of how much better it was than its predecessor. It also sparked a wave of concerns from artificial intelligence safety advocates—Elon Musk included—over the pace at which the technology was improving.

Orion is better than GPT-4, but the jump in quality pales in comparison with the one we saw between GPT-3 and GPT-4. OpenAI may even move away from its “GPT” naming convention for LLMs, which started in 2018 with the release of GPT-1.

So when Altman wrote “o1 and its successors,” he could mean that Orion will be fused with reasoning and branded as “o2.”

As One AI Scaling Law Fades, Another Takes Its Place

Let’s get back to the GPT slowdown. It may be tempting to pounce on our report and wonder if AI scaling laws—the idea that using more data and compute power in the LLM pretraining process produces better and better AI capabilities—aren’t everything they were cracked up to be. We’re wondering that too, as are Ben Horowitz and Marc Andreessen, who hold stakes in AI firms including OpenAI and recently talked about the issue on a podcast.

But that might not be the best way to think about it. The traditional AI scaling law doesn’t just mean that more data and power during the LLM training process is all you need to get a better outcome. OpenAI researchers did all kinds of other interesting things to help GPT-4 be massively better than GPT-3, such as introducing a concept known as model sparsity (read more about that here).

And making grand pronouncements about the trajectory of AI because of the GPT slowdown would ignore the company’s recent reasoning model—which stems from the now-infamous Q* breakthrough we told you about a year ago, later named Strawberry.

The reasoning paradigm could make up for the slowing improvement in pretraining results—in essence, it could represent a new type of AI scaling law. Reasoning models’ performance gets better the more time they have to think about a question before giving an answer, OpenAI has repeatedly said. It’s known as log-linear compute scaling.

Given how relatively few developers are using o1, in part because of the high price OpenAI charges for it, we can’t tell how true that is. But it’s the trillion-dollar question that’s now in front of us.

Some of you might also be thinking: if GPTs aren’t accelerating, does that mean the doomers are wrong and AI won’t enter into a so-called recursive self-improvement loop in which it figures out how to make the next, better version of itself over and over again (and then maybe conquer us all)? Andreessen, for his part, thinks the apparent plateau means such fears seem unfounded, for now.

Last but certainly not least, the data center wonks must be wondering if these shifts mean that OpenAI’s dream of a $100 billion supercomputing cluster is waning. We can’t know for sure, but all the major AI developers are full steam ahead on much-smaller-but-still-wildly-expensive clusters. (Read more about OpenAI’s upcoming cluster.)

That may be because despite the deceleration of improvements to pretrained LLMs, any improvements may well be worth the added data center cost if the resulting pretrained LLM is marginally better than those of rivals. As we point out in the piece, the better the LLM you have, the better the result you get after baking a reasoning model into the LLM. And AI developers may find large clusters particularly useful for improving these models after they are pretrained, to handle reinforcement learning during the post-training phase, and to update or tweak a model after that.

A string of busted deals has hit the venerable hedge-fund industry strategy of merger arbitrage

It has been a rough year for betting on big takeovers. Now, some in the hedge-fund industry hope President-elect Donald Trump could breathe new life into the business, by giving American corporations more freedom to pursue major deals.

The optimism is already evident in surging stock prices for some companies, like Discover Financial Services DFS 3.60%increase; green up pointing triangle, that are pursuing combinations. It is part of a broader optimism across Wall Street about potentially lucrative opportunities under a second Trump term.

In recent months, numerous high-profile acquisitions have either struggled to close or failed, including a plan to unite Tapestry and Capri Holdings, the companies behind Coach and Michael Kors. The pileup has ripped through the hedge-fund world, leading to unexpected losses and several top traders losing their jobs.

Now however, the new president’s pro-business, lighter-regulation agenda “means a pro-merger environment that will embolden companies to get bigger through acquisitions; it is as simple as that,” said Orkun Kilic, a London-based arbitrager. “I am very excited for the next four years.”

A hostile deal environment has largely kept Kilic on the sidelines since December, he said, and he now expects to become more active with Trump’s election.

Trump-related euphoria quickly fed into the shares of Capital One Financial and Discover Financial, which soared on renewed bets their $35 billion merger will pass regulatory muster.

The market is assigning an 80% probability the deal closes, up from 70% before the election, according to MKP Advisors, which specializes in M&A transactions.

A gauge of so-called merger-arbitrage performance is up just 4.6% this year through October after fees, data from the hedge-fund research firm PivotalPath shows. That compares with an 8.5% rise for the firm’s broad hedge-fund industry index.

“They’re not only underneath the industry average—I think this year they might not even beat Treasurys,” said Omar Sayed, a longtime merger-arb trader who left Millennium Management earlier this year. The yield on a low-risk 10-year U.S. Treasury note is about 4.3%.

Sayed said he plans to launch his own merger arbitrage-focused firm, but will wait until conditions improve and he can deliver the kinds of returns that investors want.

Merger arbitrage was made famous by the likes of billionaire investor John Paulson—now a candidate for Treasury secretary—and Robert Rubin, a former holder of that role and Goldman Sachs executive. It appeals partly because returns aren’t usually linked to broader market moves.

But this year, dealmaking has encountered regulatory and political pushback in Europe, the U.S. and elsewhere. In the U.S., Federal Trade Commission Chair Lina Khan has taken an unusually tough stance against mergers. FTC challenges to many deals, including Tapestry’s bid for Capri, have made it riskier to bet on takeovers closing.

Further complicating matters is a drop-off in deal numbers, which hit their lowest level since 2015 in the first 10 months of this year, LSEG data shows. And higher interest rates have helped make it more expensive to use debt to boost deal returns.

Trump’s credentials as pro-merger will be tested early with his decision over the fate of Khan, the FTC chair. Many investors expect her removal from this role because of her record of challenging many deals and wariness of big tech.

This year has brought a string of busted or challenged mergers, including Amazon.com’s abortive takeover of robot-vacuum maker iRobot, and a failed $3.8 billion union between JetBlue Airways and Spirit Airlines.

This rough ride harks back to 2014 when merger arbs suffered a one-day shellacking some dubbed “Arbageddon,” as two huge deals collapsed and President Barack Obama said his administration would revisit rules on tax-driven mergers.

Last month, merger investors suffered two major setbacks. The first drubbing resulted from the collapse of a bid for Hong Kong-listed China Traditional Chinese Medicine Holdings, after the $2 billion deal unexpectedly failed to win timely approval from Chinese regulators.

Days later, a U.S. judge blocked Tapestry’s bid for Capri, siding with the FTC and arguing the deal would raise prices for more-affordable handbags. The companies plan to appeal. Shares in China TCM and Capri suffered one-day drops of 35% and 49%, respectively, after the deals collapsed.

Izzy Englander’s Millennium Management ranked among China TCM’s biggest shareholders before the deal broke down, owning a more-than-5% stake worth about $135 million, filings show. Hong Kong-based Athos Capital owned more than 4%, another filing showed.

Millennium was also a shareholder in Capri, as was London-based Melqart Asset Management, FactSet data as of June shows. The firms each owned more than 2 million shares at the time.

The setbacks have left some money managers struggling to hang on to their jobs.

At least 10 merger-arb investors and analysts since April have been fired, left or are planning to leave their jobs following underperformance, according to people familiar with the matter.

Among those who have departed their roles: Ed Cooper and Laurent Pujade-Lauraine at Millennium Management, Eric Fritz at Schonfeld Strategic Advisors, Adam Schwartz at Balyasny Asset Management and Neel Gupta, who worked at Soros Fund Management.

The managers and their firms either declined to comment or couldn’t be reached for comment.

The Tapestry-Capri deal had been a popular target for merger arbs. Manuel Blanco, a portfolio manager at Maven Securities, touted the deal’s merits at a June conference. That included an option strategy to mitigate potential losses if the deal failed, and the possibility of still making money if that scenario led to a breakup of Capri, according to a copy of his presentation seen by The Wall Street Journal.

Blanco is in the process of leaving Maven. He said his departure has nothing to do with his Capri investment and is driven by his plans to set up a new fund.

Despite the changing atmosphere, there is one big deal that many traders view as a lost cause even under a Trump administration: Nippon Steel’s high-profile and controversial $14.1 billion deal for United States Steel. Trump came out against the tie-up during the campaign, amid opposition to the Japanese takeover from the United Steelworkers union and local politicians.

Latest fundraising values company at $2.75bn following one of biggest VC rounds in Latin America in recent years

Argentine fintech Ualá has raised $300mn in one of the largest fundraising rounds for a start-up in Latin America in recent years, in what libertarian President Javier Milei described as “proof” of an impending wave of tech sector growth in the notoriously tumultuous economy.

The latest fundraising takes Ualá, which launched in 2017 and now has 8mn customers and banking licences in Argentina, Mexico and Colombia, to a valuation of $2.75bn.

The round is one of the largest in the region in recent years, according to data from Dealogic, following a $500mn fundraising for cold-chain storage firm Emergent Cold LatAm in 2023. Brazilian fintech Neon Pagamentos also raised $300mn in 2022.

The round was led by Allianz X, the growth investment arm of insurer and asset manager Allianz, and also included hedge fund manager Alan Howard. Ualá is backed by the SoftBank Latin America Fund and China’s Tencent.

The move follows three years of relatively scant venture capital investment in Latin America amid higher US interest rates. 

Milei, who is 11 months into a drastic austerity and deregulation drive that has slashed Argentina’s sky-high inflation rate, said on Monday that Ualá’s fundraising was “proof that . . . Argentina has everything it needs to become a technological power”.

“We are a pressure cooker ready to boil over with potential,” he added.

Argentina has one of the largest tech industries in the region and is the birthplace of ecommerce giant MercadoLibre and software development firm Globant. The sector stands to be a big beneficiary of Milei’s plans to stabilise the macroeconomy, scrap regulation and eventually lift currency and capital controls that have been in place for nine of the past 13 years, which have deterred investment.

“This fundraising round will allow Ualá to never raise capital again and to reach full profitability in all our markets, after already achieving that in Argentina this year,” Ualá founder and chief executive Pierpaolo Barbieri told the Financial Times.

“Our goal with this round is to become the largest bank in Argentina, period,” he added. 

Ualá has grown rapidly in Argentina by capitalising on the country’s bureaucratic banking sector and large unbanked population. Almost a fifth of Argentine adults now hold a Ualá account, according to the company, which offers debit and credit cards, payment services, loans and investment products. 

The firm had quintupled its credit book in 2024, Barbieri said, as Milei’s reform drive begins to expand Argentina’s long-stunted private credit market, which has shrivelled further in recent years amid a severe economic crisis.

Argentina’s private credit market currently measures just 5 per cent of GDP compared with more than 50 per cent in Brazil and 90 per cent in Chile, Barbieri said.  

“It should at the very least, quadruple to match where it was in 2017, 2018. But then it should grow by 10 or 20 times [over] a couple of decades,” he added. “The direction of travel is very clear — if the macro stabilisation is successful.”

Broker dealer previously claimed it had been victim of ‘massive fraud’ by Michael Viney and another former employee

Howard Lutnick’s BGC Group has brought contempt of court proceedings against a former UK employee, accusing him of hiding assets after he took part in a £24mn scheme to defraud the US broker dealer.

Lawyers acting for the company told the High Court in London that tax adviser Michael Viney had breached an earlier freezing order on his assets, including by selling a property in Barcelona without disclosing it.

BGC, which is led by billionaire Lutnick — one of Donald Trump’s closest allies and co-chair of his transition team — previously claimed that it had been the victim of a “massive fraud” perpetrated by Viney and another UK employee, Xavier Alcan.

The company reached a settlement with Alcan earlier this year. Edward Levey KC, representing BGC, told the court on Monday that Alcan “defended the matter right up to the eve of trial” but the claim was then settled on terms that he pay “significant sums” to BGC.

Viney has previously admitted he was liable to pay BGC. According to the company, he was told to pay about £24mn under the terms of a court order in May 2022. It served a contempt of court application on him in June.

Levey claimed on Monday that there was “overwhelming evidence” that Viney was in breach of a court injunction.

In written submissions, he said that Viney had sold a property in Barcelona without BGC’s knowledge and also “failed to disclose” a £1.26mn investment.

The company had previously accused Viney and Alcan of diverting payments that were intended to be between BGC and other legitimate sources, mainly HMRC, to bank accounts in numerous jurisdictions.

BGC said Viney used a “trusted position” within the company to conceal 32 separate fraudulent payments, including by doctoring emails. Some of the proceeds had been used to buy properties, jewellery and other luxury items, the group claimed.

A hearing on Monday was adjourned until the new year as Viney lacked legal representation.

Viney, who appeared before the judge without counsel, was not asked for his substantive response to the contempt application. He said he was making his best efforts to secure legal aid solicitors.

In written submissions, Levey claimed that the Barcelona property had been sold for €253,800 and that the proceeds were paid into a previously undisclosed bank account. Some of the money had been recovered but a large chunk had not been, he maintained.

Levey also contended that there had been a “dissipation” of income generated from rental properties in the UK and Florida and that Viney had been “planning to get his hands” on an undisclosed investment and would have done so had BGC “not put a stop to it when it learned of its existence”.

Lutnick, who also heads Wall Street firm Cantor Fitzgerald, personally gave more than $10mn to Trump’s re-election campaign.

From self-driving cars to AI that writes enterprise software: Cogna founder raises $15M

A founder who was an early mover in the race to build autonomous vehicles has raised $15 million for his next act: a startup that claims its AI can write enterprise software on its own.

Cogna — as the U.K.-based startup is called — is led by Ben Peters, the technical co-founder of FiveAI (the self-driving startup that was acquired by Bosch in 2022). Notion Capital is leading the Series A, with Hoxton Ventures and Chalfen Ventures also participating. It comes on the heels of Cogna, founded in May 2023, raising a seed round of $4.75 million earlier this year from a number of investors, including Peters’ FiveAI co-founder Stan Boland and Herman Hauser, the founder of Acorn Software.

The focus of Cogna is the world of enterprise resource planning. ERP is a dry but very necessary software component in the running of organizations, covering everything from procurement and supply chain and inventory management through to risk assessments, finance and human resources. Typically large enterprises can pay up to billions of dollars in contracts with systems integrators and consultancies to handle their ERP, either by customizing off-the-shelf software for clients, or writing custom applications from the ground up, to fit an organization’s particular needs.

In typical AI startup fashion, Peters believes that this work is best done, and can be compressed into, an AI platform built for the purpose. While some challenges have been significantly trickier for AI to address than others, there have been some early signs of ERP potentially being one area where it might stick, and for Cogna to potentially become one of the players delivering on that pitch. A year since launch, Cogna has signed up customers that include the U.K. gas distributor Cadent Gas and infrastructure and utilities service provider Network Plus.

“To be clear, none of our customers think of it as ERP software,” Peters said in an interview. “For them, they’ve got a problem which isn’t solved by using SAP or [another] legacy system. We deliver a custom, precision-built piece of software for that. They experience [it] ike a classic piece of SaaS, but it happens to have been built specifically for them, specifically for their workflow. And it’s written by our AI.”

Cogna balances a mix of generative AI with other kinds of tooling to actually put together its custom software. Non-technical teams can describe their pain points in natural language, and the Cogna does the rest, the company claims.

“We’re a team of experts in domain-specific languages, compilers, AI and reliable and scaleable enterprise SaaS,” Peters said.

StrictlyVC San Francisco

Mix and mingle with your peers and hear insights from top-tier VCs

San Francisco | December 4

REGISTER NOW

These days, it’s not uncommon to come across AI startups leveraging multiple large language models simultaneously depending on a particular task at hand — it’s the customization that makes a startup unique — and that is the case here, too. Peters said that the language models used for people to interact with its platform come from a variety of providers that include OpenAI and Anthropic. These are “a key part of our ‘Natural Language Compiler,” Peters told TechCrunch over a call.

On top of that, Cogna is building an engine so that “the software can truly write itself,” said Hussein Kanji, the co-founder of Hoxton Ventures.

In doing so, Cogna is an example of how generative AI — which saw a viral explosion with the launch of consumer-accessible services like ChatGPT — is indeed being folded into enterprise usage and more complex applications. That is important because a number of foundational AI companies, like OpenAI, have made it clear that they, too, have ambitions to tap into the enterprise market, a highly lucrative area that will help them generate the vast revenues that investors expect.

Kanji likens what Cogna is doing to Cursor, the code building tool from OpenAI-backed Anysphere. We broke the news last week that Anysphere is in the middle a massive, unsolicited bidding war among investors that want to back it, valuing Anysphere as high as $2.5 billion on the back of massive business growth, so the area is clearly very hot.

“That’s kind of what Cogna is also doing,” he said.

Based on his’ experience at self-driving tech company FiveAI, Peters is adept at spotting openings in the market early, but he also has experience in how to shift gears, so to speak, when plans don’t go as you expect them to.

“We had cars driving around London in 2018,” he recalled. “[But] we then pivoted to providing the development environment to develop a stack for other companies that were building their own electronic design automation stack. We were running a million simulations a day by the time we sold the company.”

Five.ai, which had raised around $78 million in funding was valued at $216 million in its last round before it was acquired by Bosch for an undisclosed sum.

Cogna now plans to go up against classic IT consultancies such as Wipro and CapGemini.

The investment into Cogna is the first from Bryan Gartner, formerly of Khosla Ventures, who joined Hoxton as its newest partner last year.

London-based portfolio manager worked on some of activist’s most high-profile campaigns

Ken Griffin’s hedge fund Citadel has hired Nabeel Bhanji, a longstanding London-based portfolio manager at Elliott Management who worked on some of the activist hedge fund’s most high-profile campaigns.

Bhanji, who worked at Elliott for more than a decade and was one of 11 equity partners globally, is expected to have management responsibilities at Citadel in addition to an investment role, according to two people familiar with the situation.

Citadel and Elliott declined to comment.

The move is the latest and most significant departure from Elliott’s London office. Paul Singer’s $70bn hedge fund had only made Bhanji an equity partner in the past year. 

Bhanji was in charge of some of Elliott’s biggest activist positions in recent years, including at miner Anglo American and Japanese technology conglomerate SoftBank. 

Citadel runs $64bn in assets under management and, since it was founded in 1990, has grown into the most profitable investment firm in the industry’s history.

Europe still has decades of under-investment in security to address

The election of Donald Trump should, on the face of it, mean a direct hit for the European defence sector. The president-elect has never been a fan of Nato, grumbling that the US pays too much, and he wants to broker a speedy end to the war in Ukraine. But investors appear unfazed. Share prices of big European defence groups continue to climb following last week’s US election.

That is less counterintuitive than it sounds. True, Trump’s approach suggests there will be fewer US dollars streaming into security and non-US military industry. The US, together with Germany, provides the biggest slice — 16 per cent apiece — of Nato’s annual running costs. It is also the biggest spender on defence, with a planned budget of $850bn. But for European defence investors, props remain.

First, creeping protectionism is not limited to the US. Europe, too, wants more spending kept at home. More soft infrastructure is in place to do so. In March, the bloc penned its first European defence industrial strategy with the aim of reversing overseas spending: 78 per cent of EU states’ defence orders from February 2022 to June 2023 went to companies outside the region.

The EU is on the verge of appointing its first defence commissioner. Andrius Kubilius, the former Lithuanian prime minister, said in last week’s hearing that he wanted to create “a true single market for defence”.

Nato’s European members are also increasing their spend. This year, 23 of 32 members are expected to meet or exceed the 2 per cent of GDP target; together, European allies and Canada will spend a combined 2.02 per cent of GDP, up from 1.43 per cent a decade ago.

US defence spending will not entirely be out of reach, either. Germany’s Rheinmetall and its ilk have US businesses and plants that may give them continuing access.

Second, despite recently increased spending on tanks and shells, Europe has decades of under-investment in security to address. Over the past 30 years, it has spent a cumulative $1.4tn less than required to meet Nato criteria, according to Berenberg estimates. This year’s $23bn overspend relative to the target is a drop in the ocean. That suggests a huge programme of equipment upgrades will buoy spending even absent war.

Manufacturers are still minting money. Prodigious defence spending sent Rheinmetall’s sales up 39 per cent year on year to €2.45bn in the third quarter. Full-year guidance is for €10bn of sales on a 15 per cent operating margin.

The real bottleneck to the European defence industry is physical: ropey supply chains and a highly fragmented procurement base. Changes in the White House make fixing that all the more important.