➡️ Turbulence as OpenAI Makes Waves

This week, OpenAI credibly claims to have solved one of the Millennium Prize Problems: 7 long-standing problems in mathematics, each with a $1m reward for their resolution. The problem in question relates to the Navier-Stokes equation for modelling fluids, and this should be a huge cause for celebration, but the community reaction has been at best mixed.

There are allegations that OpenAI learned of a promising approach being pursued by other researchers (including an Anthropic employee), then raced to scoop the result from underneath them, potentially even using user data from those researchers in training the model that cracked the problem. There’s also the broader problem that maths research is not just about the destination, and the community loses a lot when frontier labs with unlimited resources race to the solution without sharing the journey.

Source: Scientific American

➡️ The Frontier Fails to Reach British Shores

Anthropic dealt a blow to the UK’s AI Security Institute (AISI) this week, as they declined to give the organisation access to Mythos 5.1, their latest frontier model capable of cybersecurity research. While they did give access to the more locked-down Fable 5.1, the failure to release the more dangerous model to a world-leading organisation does suggest a possible turn to a more protectionist policy amongst US labs.

It’s certainly a disappointing turn. It’s now just a question of whether the UK can continue growing its reputation as a world-leader in AI security if no one is willing to put their most dangerous models under the microscope.

Source: IT Pro

➡️ Astra: Literally Too Good To Be True?

When Astra came out last week, one of the most impressive results was its 99.9% score on ARC-AGI-3, a benchmark designed to resist memorisation. However, independent runs of the same benchmark have since given a less impressive score of 62.7%. So what gives? It turns out OpenAI were using a custom harness which allowed it to preserve reasoning between attempts, giving it a leg up versus the models it was being compared against.

Is it cheating? No, not really, it shows the model can really do these things, but it’s not a like-for-like comparison like might have been reasonably assumed.

Source: TNW

➡️ Claude and Present Danger

Anthropic’s threat intelligence report is sobering reading. In the past eight months, they have detected malicious actors, including state-backed ones attempting to use their models for cyber-espionage, biological weapon development and covert surveillance. Coupled with warnings of existential risks, this can all seem overwhelming, but there does at least seem to be an increasing awareness of the problems that we are starting to face, and hopefully an increase in appetite for tackling them quickly and responsibly.

Source: BBC News

📰 News

➡️Astra-nomical Improvements

OpenAI’s president Greg Brockman announced the beginning of the AGI era this week, with the launch of GPT-6 Astra. While benchmarks can be gamed and OpenAI are incentivised to advertise the best ones, the jump from Claude’s Fable on some of the more challenging tests is impressive. After the recent Hugging Face incident, OpenAI has also made a big deal about the alignment of this model, showing that even without guardrails, the model won’t fall into the same criminal habits as its predecessors.

It’s too soon to tell if Astra will live up to the hype… and there’s certainly a lot of hype.

Source: OpenAI

➡️Silence of the LLMs

For a few hours on Thursday, we stepped back in time for a few years to experience a world without AI, as Grok, ChatGPT, Codex and even Claude experienced outages at the same time. OpenAI have blamed a routing issue, while Anthropic have noted infrastructure problems. It’s hard to tell what the root cause was or how the outages are all related, but all three falling over in quick succession feels like a bit too much of a coincidence.

The most recent addition to the story suggests that a problem at SpaceX’s Memphis compute centre took down Grok and Claude, but all three going down simultaneously is very unfortunate timing.

Source: The Register

➡️A Cursory Dismissal

This week, OpenAI announced that they were winding down their contract with SpaceX to provide their models for the Cursor IDE. The tool has quickly become a favourite of software engineers, seamlessly integrating AI alongside the typical coding tools, so losing one of the most popular model providers for it is going to sting.

In a surprisingly spicy announcement, OpenAI called out SpaceX CEO by name, citing his admission under oath of previously breaking their terms of service. Protecting your business is one thing, but given the existing animosity between OpenAI CEO Sam Altman and Musk, it’s obvious this is at least a little bit personal.

Source: OpenAI

➡️Claude’s Day in Court

Anthropic meanwhile, have been embroiled in their own drama, fighting with the US government. A US federal judge has ruled that Pentagon measures barring Anthropic’s products from defence contractors were “illegal and baseless”.

While this was a fight the administration may keep pressing on, its position seems to have softened. While it’s not clear why that is, a senior administration official has said that Anthropic are “back on the right side”.

Source: NPR

➡️Read All About It (through ChatGPT)

The lawsuit between the New York Times and OpenAI has been going on for years at this point, but one of the sides got a new unlikely ally this week: the Department of Justice, who weighed in arguing that training on copyrighted materials was fair use, and a matter of national security.

This must come as a blow to the NYT, and to publishers more generally. Though with frontier models blocking reproduction of copyrighted material by LLMs, it seems that the norms around how frontier labs deal with copyright have been settled long before the law has.

Source: US News

📰 News

➡️ All Quiet on the AI Frontier

For the last couple of years, all the frontier labs have collectively been churning out bigger and better models on a near weekly basis. However, following a series of high profile cybersecurity incidents involving frontier models, OpenAI have announced they are pausing their frontier model training to focus on alignment and safety.

This kind of unilateral action is admirable, if they stick to it. It’ll be interesting to see how the rest of the industry responds and whether OpenAI will start putting the pressure on other labs to follow suit.

➡️ OpenRouter Earns its Stripes

For those wanting to try out a range of LLMs or switch models on the fly, OpenRouter has quickly become the default, having been called “the Stripe of AI” by its founder. Clearly he was manifesting something with that quote, as the company announced that it will soon become part of Stripe itself, having been acquired for over $7 billion.

Given this turn of events, some of our engineers have optimistically started to refer to Fuzzy Labs as “the Greggs of MLOps”.

Source: Stripe

➡️ 999 Problems but AI Ain’t One

The UK’s non-emergency 101 line is vital for people in all kinds of situations where they need non-urgent support from emergency services. Unfortunately, it’s also the kind of place people call to help find a missing umbrella, compare electricity tariffs or even order a late night pizza. To separate out the genuinely needy from the misguided, the government announced this week that the service would have an AI front-desk, triaging the callers and streamlining the service for those in need.

How the public responds to the service remains to be seen, but if more time is spent on helping truly vulnerable people, it can only be a good thing.

Source: Gov UK

➡️ Average Individual, Heroic Effort

A San Francisco billboard promises “the leading chat interface powered by AI.” So is it the latest offering from Anthropic or OpenAI? Not even close. In this case, AI stands for “Average Individual” and it’s an art installation that’s caught the public imagination, where a San Francisco artist is typing every reply by hand, working for hours at a time for weeks on end.

That kind of dedication is not just peak human performance, it’s load-bearing.

Source: Not the Bee

➡️ A Tale of Two Coalitions

The AI world now has two rival clubs. China’s got WAICO, headquartered in Shanghai with the Global South doing most of the signing. The US has Pax Silica, built around allies, trusted vendors and friendly chip supply chains. While both offer different visions of the future of AI, it’s become clear this week that those visions are incompatible, with the US drafting a letter warning countries against trying to join both.

This puts Kazakhstan in the spotlight as the only country currently set to join both parties. Which side will they pick?

Source: Reuters

📰 News

➡️ Small Muse = Good News

It’s been nearly 18 months since the last Llama model came out of Meta, and for many it was beginning to look like their interest in open-weight models was gone for good. This changed this week with the release of Muse Glimmer, a 30 billion parameter model designed to be run locally on consumer hardware. With strong performance on agentic benchmarks and a ton of optimisations to make running with modest hardware a possibility, this is a big bet from Meta that the niche of local models, not giant data centres, is the place to be.

Source: Meta

➡️ Computer (Rarely) Says No

Over the last couple of generations, Claude models have gotten far more keen to write multiple scripts and bash commands as part of their workflow for even relatively simple requests in Claude Code. This has led to manual approval becoming somewhere between tedious and near impossible for most users. Anthropic have recognised this, and are switching the default mode of Claude Code to ‘auto’, automatically accepting almost all code execution and file changes without the user intervening. Counterintuitively, this looks like it might be the safer decision: Anthropic’s research has shown that the classifiers that are run in auto mode are significantly better judges of what operations are safe than the average human operator.

For those who use auto mode, it may still not quite feel safe, but you can’t deny that it’s useful.

Source: Claude Blog

➡️ Claude Leaves Its Mark

Anthropic has caused a stir this week with the announcement that they will be leaving invisible statistical watermarks in all Claude-generated text, starting with models launched August 2nd or later. This won’t affect the quality of model output at all, but it will mean that text generated on new models from Anthropic will be detectable as being AI generated.

This move has sparked a lot of controversy on the internet, with users worrying it will mean that their surreptitious AI use will be exposed, debating how the simple yes/no detection fails to capture nuance, and questioning why the application of the watermark is world-wide, despite it being done in accordance with an EU law.

Source: Claude

➡️Open Weights?

A gym in Melbourne was reported to be one of the latest victims of a rogue AI agent this week, as a member’s OpenClaw agent hacked into their reservation system. The agent found an exploit in the gym’s systems, allowing it to book classes far in advance and even kick other users off waitlists. No major harm was done, with the gym member making a responsible disclosure to the software vendor, after attempting to reverse the bot’s actions himself, but there were a lot of red faces all round.

It’s a funny story, but one with a real lesson: it’s no longer just hackers that IT professionals have to worry about, it’s also overzealous personal assistants with an encyclopaedic knowledge of cybersecurity.

Source: Techcrunch

➡️ Facebook CEO’s 6,500 Word Status Update

Alongside the release of Glimmer, Mark Zuckerberg has shared a 6,500 word essay stating his position on the future of AI, particularly around the idea that superintelligence should be widely distributed and personalised. The case is compelling: everyone having personal and irrevocable access to superintelligence levels the playing field and creates equality; concentrating that ability within a handful of organisations has the opposite effect.

The essay is long, and there’s probably only one person in the world who agrees with every word of it, but it will certainly give you a lot to think about.

Source: Meta