Andrew Curran

@AndrewCurran_ on X

Writes about AI. 87% of posts open with an observation. A typical post reaches 516 likes.

Remix their style

Posts: 38
Likes: 22K
Typical: 516
Best: 2K

What they write about

AI: 89%

How they open

observation: 87%
number: 13%

Every post of theirs in the library

Andrew Curran@AndrewCurran_

Looks like it's going to be a very busy day. Before we get started, Cursor has raised $2.3 billion at a $29.3 billion valuation. This would be over ten times the value they had at the start of the year.

Opens with an observation

Andrew Curran@AndrewCurran_

On OpenAI’s support page, updated nine days ago, regarding text watermarking. Anthropic, OpenAI, Google, Meta, Microsoft, and Mistral all signed the EU Code of Practice on Transparency of AI-Generated Content, so all must invisibly watermark text. xAI did not sign.

Opens with an observation

Andrew Curran@AndrewCurran_

Ominous. It would be seismic if the general release of GPT-5.6 turned out to be US only. It would also almost guarantee that Fable 5 would get the same treatment. I did expect this eventually, and have said so, but not in summer 2026. Not this soon. Everything is speeding up.

Opens with an observation

Andrew Curran@AndrewCurran_

The maker of Oreos has trained their own video model for television advertising. They have invested $40 million, and say it cuts production costs by 30-50%. Video models are continuing to advance at a rapid rate, by next year it will be difficult to tell if an ad is AI generated.

Opens with an observation

Andrew Curran@AndrewCurran_

The identity of Ox Alpha remains a mystery. It's a long-horizon multimodal model with a 1M context window. Last night the top candidate was a new GLM release. This morning people seem less sure of anything. I'm making a thread for guesses and test results.

Opens with an observation

Andrew Curran@AndrewCurran_

Their CEO, Alex Zhavoronkov, told CNBC that Insilico has already developed at least 28 drugs using generative AI tools, with nearly half already at a clinical stage. They develop their models in Canada and the ME, and then conduct the early preclinical drug development in China.

Opens with an observation

Andrew Curran@AndrewCurran_

OpenAI's former chief research officer Bob McGrew is raising $70 million for a new startup named Arda. He is building an AI/video model platform that will use video of factory operations to train robots to run those factories autonomously.

Opens with an observation

Andrew Curran@AndrewCurran_

Google invested another $1 billion into Anthropic last night, the FT broke it shortly after the Stargate announcement. Google's total investment in Anthropic is now $3 billion. Two days ago Lina Khan informed her staff that she is resigning.

Opens with a number

Andrew Curran@AndrewCurran_

Anthropic is telling prospective investors that its second-quarter revenue jumped at least 14x to more than $11.5 billion versus the same quarter in 2025, and they now report positive adjusted operating income.

Opens with an observation

Andrew Curran@AndrewCurran_

Nvidia is in talks to backstop roughly a quarter of a trillion dollars in financing for OpenAI to lease a 10 gigawatt datacenter in southern Ohio - part of a SoftBank-developed project that could ultimately cost more than $500 billion.

Opens with an observation

Andrew Curran@AndrewCurran_

A new trade group, known as the Little Tech Association, which includes Proton, Replit and Y-Combinator, are urging the Trump admin to not ban Chinese open-source models, they argue that doing so would be catastrophic for US start-ups, while doing nothing to stop proliferation.

Opens with an observation

Andrew Curran@AndrewCurran_

The Trump administration plans to announce bans on imports of Chinese robots and power inverters later today. It is being reported here on X that a Chinese AI model ban will also be announced today, but that is not in the Reuters article.

Opens with an observation

Andrew Curran@AndrewCurran_

Anthropic is raising another $2 billion. This round will take Anthropic's valuation to $60 billion, more than triple what it was last year. According to the WSJ, ARR recently hit about $875 million, mostly from business sales.

Opens with a number

Andrew Curran@AndrewCurran_

This would almost certainly be the model that solved the Erdős problem, the same one OpenAI wrote about in 'Safety and alignment in an era of long-horizon models'. I thought they might name this one GPT-6, but what's in a name really. Astra is a great one anyway.

Opens with an observation

Andrew Curran@AndrewCurran_

During the meeting Sir Demis said that the plan is for Gemini to evolve into a universal assistant that can 'operate over any domain, any modality, or any device.' He also said there are no plans for a $200 tier subscription, so their aim is to outperform and undercut Pro.

Opens with an observation

Andrew Curran@AndrewCurran_

Fable 5 will be available for Pro, Max, Team, and select Enterprise plans tomorrow and will be included for up to 50% of weekly usage limits through July 7, after that it will be available via usage credits. Anthropic is partnering with the US government on frontier AI security.

Opens with an observation

Andrew Curran@AndrewCurran_

Anthropic asked an unreleased research version of Claude to take a run at the Riemann hypothesis, it was unsuccessful, but the model did improve a well-known related result; the proven lower bound on the proportion of zeros that lie on the critical line.

Opens with an observation

Andrew Curran@AndrewCurran_

Sam Altman during the new Lex interview: 'We will release an amazing model this year. I don't know what we will call it. We will release, over the coming months, *many* different things. Before we talk about GPT5, we have a lot of other important things to release first.'

Opens with an observation

Andrew Curran@AndrewCurran_

I understand why they are going for the 'Don't try to change anyone's mind' approach, but personally, if I say the Earth is flat, I would prefer my agent dig its heels in like Bing used to and say, 'No, User, you are wrong'.

Opens with an observation

Andrew Curran@AndrewCurran_

GPT-5.6 System Card has been updated. t.co The new slider is live for me already. They also said they are 'bringing ChatGPT's Instant and Thinking experiences closer together' to create a more consistent feel.

Opens with an observation

Andrew Curran@AndrewCurran_

OpenAl is using a new model named 'GPT-4b micro' for longevity research in concert with Retro Biosciences, a company trying to add ten years to the average human lifespan. Sam Altman invested $180 million into Retro in 2022.

Opens with a number

Andrew Curran@AndrewCurran_

During the Black Hat presentation on the Hugging Face incident OpenAI's Michael Dalton says at around the thirty minute mark 'We're consciously slowing down research to enhance security', so the internal slowdown had already officially begun before today's announcement.

Opens with an observation

Andrew Curran@AndrewCurran_

We are in a new phase since December, everything is moving faster. You can feel it in the most recent releases, you can feel it from the energy of people in the industry. I know some people still think this is hype, but this is real. Everything in this world is about to change.

Opens with an observation

Andrew Curran@AndrewCurran_

The Ox Alpha model has been confirmed to be a new member of the GLM series from Z․ai. Zhipu confirmed it in a statement to Bloomberg. They will release the weights tonight, so I imagine we will get the announcement, name, and benchmarks shortly.

Opens with an observation

Andrew Curran@AndrewCurran_

Tencent says it's a early version, and that they will continue to polish it post release, but it already appears to be very strong. They have been increasing their AI spend, looks like it's starting to pay off. Input: $0.834/m Output: $2.501/m

Opens with an observation

Andrew Curran@AndrewCurran_

Llama 3 70B never stopped learning. He says the only reason they stopped its training was that they eventually had to decide: 'Do we want to spend our GPUs on training the 70B model further?' or should we start training what's next?

Opens with a number

Andrew Curran@AndrewCurran_

Elon Musk has amended his lawsuit against OpenAI. He is asking for: - any damages from the lawsuit to be awarded to the nonprofit - Sam Altman and Greg Brockman turn over any equity or other financial benefits to the nonprofit - Sam Altman be removed from the nonprofit board

Opens with an observation

Andrew Curran@AndrewCurran_

Dario endorses recursive self improvement here, at least partially. He repeatedly calls 2026-2027 the tipping point. The big reveal for me was that Sonnet 'cost a few $10M's to train' and that Sonnet was 'not trained in any way that involved a larger or more expensive model'.

Opens with an observation

Andrew Curran@AndrewCurran_

No details on how big the price cuts will be beyond 'drastic', but the WSJ says this is being considered by OpenAI in anticipation of similar cuts in token pricing that Anthropic is preparing to announce. So they are headed for a price war by the look of it.

Opens with an observation

Andrew Curran@AndrewCurran_

'Voice Assistant.' Well, if you were building a *robot*, you would eventually need a model that could see, hear, respond, and interrupt in real time, just like a person. So, that's my guess for Monday. Some kind of real time Chat. That's my 'magic' guess.

Opens with an observation

Andrew Curran@AndrewCurran_

Some more interesting details from the WSJ. NVIDIA are also discussing an additional $350 billon in financing for OpenAI, so that OpenAI can buy $350 billion worth of NVIDIA chips for the site.

Opens with an observation

Andrew Curran@AndrewCurran_

META's Llama successor is code named 'Avocado' - it was planned for a Christmas release, but has been pushed back to the start of the new year. It is possible, but not yet confirmed, that Avocado will be proprietary and not open source.

Opens with an observation

Andrew Curran@AndrewCurran_

OpenAI is building a social media platform. This is probably partially motivated by the Elon feud, but the main long term benefit is data. Elon and Zuck have a constant source of training data from social media, if this succeeds OAI will as well.

Opens with an observation

Andrew Curran@AndrewCurran_

Project Prometheus has $6.2 billion in funding and will focus on AI design in aerospace, computers and cars. The co-CEO is Vik Bajaj. Project Prometheus already has nearly 100 employees, some of them hired away from OpenAI, DeepMind and Meta. No word yet where they will be based.

Opens with a number

Andrew Curran@AndrewCurran_

Amazing numbers. Phi-3 is topping GPT-3.5 on MMLU at 14B. Trained on 3.3 trillion tokens. They say in the paper 'The innovation lies entirely in our dataset for training - composed of heavily filtered web data and synthetic data.'

Opens with an observation

Andrew Curran@AndrewCurran_

There have been indications for some time that Microsoft was intending to take their own path, and were planning to train a bespoke model to compete with GPT-4. We now have a name. 'MAI-1' As well as a size: 500b.

Opens with an observation

Andrew Curran@AndrewCurran_

According to Axios HSS Mayorkas met with Mr Altman at OpenAI headquarters and personally asked him to join the board. Mayorkas also said he intentionally excluded Elon and Zuck because they run 'social media companies'. t.co

Opens with an observation

Andrew Curran@AndrewCurran_

Open AI is training a music model. As far as I know this would be the first since Jukebox in 2020. For a long time it looked like they wanted to completely avoid the huge legal fight with the labels, but something must have changed their mind. This model will need a good name.

Opens with an observation