Jim Hawkins
LightNews LightNews
banner
jimallthetime.bsky.social
Jim Hawkins
@jimallthetime.bsky.social
2.8K followers 4.5K following 6.8K posts
Podcaster at Hello You the Shropshire Podcast (jimallthetime.substack.com). Charity worker. Ex BBC radio presenter/journo

Also a photographer, musician, MC at Shrewsbury Folk Festival. Good listener, empath, creative. He/him. Say hello?
Posts Replies Media Videos
Pinned
jimallthetime.bsky.social
Jim Hawkins @jimallthetime.bsky.social · Nov 1
Thrilled and honoured by this appreciative comment on the most recent episode of Hello You the Shropshire podcast - The Lost Art of Noticing - by the super-noticer herself, @mirandakeeling.com 🥰

Hope you enjoy it as much as Miranda did!

jimallthetime.substack.com/p/hello-you-...
jimallthetime.bsky.social
The President is Unwell

Donald Trump's response to Rob Reiner's murder proves he is unfit to lead

open.substack.com/pub/lizplank...
The President is Unwell
Donald Trump's response to Rob Reiner's murder proves he is unfit to lead.
open.substack.com
December 15, 2025 at 7:05 PM
Reposted by Jim Hawkins
edzitron.com
Another myth is that Anthropic is "more efficient than OpenAI" and will "break even." Anthropic raised $16.5bn compared to OpenAI's $18.3bn in 2025, and burns billions of dollars too. Both are horribly unprofitable with no path to sustainability.
www.wheresyoured.at/ai-mythbuste...
“Anthropic Is More Efficient Than OpenAI” In A Way That Makes It Somehow Profitable In The Future I simply do not think this is true. Per Fortune’s recent Anthropic piece (emphasis mine): Even more remarkably, Anthropic is generating such growth without spending nearly as much as some rivals—at a time when massive capital expenditures across the industry are stoking anxiety about an AI bubble. (OpenAI alone has signed AI infrastructure deals worth more than $1 trillion.) That’s in part because Anthropic says it has found ways to train and run its AI models more efficiently. To be sure, Anthropic is nowhere near profitable today: It was pacing to end 2025 having consumed $2.8 billion more cash than it took in, according to recent news accounts citing forecasts provided to investors. But the company is also on track to break even in 2028, according to those projections—two years ahead of OpenAI. Let me lay some things out: I estimate that Anthropic will make $4bn to $5bn this year, so around half of what OpenAI makes (I estimate OpenAI will make $8bn to $10bn, max). Based on my reporting on its AWS bills, Anthropic spent $2.66 billion through the end of September, a month where it spent $519 billion. As a result, it’s fair to assume that it spent that much in the remaining three months of the year (if not more!), bringing its AWS bill up to about $4.2 billion. If we assume that Anthropic spends about as much on Google cloud, that brings its compute costs to $8.4 billion. Assuming Anthropic makes $4.5 billion in revenue, that means that Anthropic spends about $1.87 per dollar of revenue just on compute. Though OpenAI’s spend is likely much higher — I’d peg its inference costs at $12 billion this year and, on the low end, its training costs at $10 billion or more — it’s likely spending $2.75 in compute per dollar it makes, which is worse, but doesn’t make Anthropic’s math better. This “efficiency” story also doesn’t make sense based on Anthropic’s funding. OpenAI has raised $18.3 billion this year for what is supposedly a “much less efficient” operation - Anthropic raised $13 billion and $3.5 billion for a total of $16.5 billion. If it were “so much more efficient,” why would that be the case? Anthropic has done a really great job taking advantage of reporters that don’t want to think too hard or consider its math, but the most egregious thing I’ve read is that Anthropic will “only burn $2.8 billion” this year. Considering Anthropic has now promised $21 billion to Broadcom for Google TPUs, the argument might be that “Anthropic needs to pay for those chips,” but if that were the case, Anthropic would have to raise far, far more. That, and Broadcom CEO Hock Tan made it clear that the backlog of revenue wouldn’t start coming in until 2027. I think the truth is simple: Anthropic is as ugly and nasty and disorganized and costly to run as OpenAI. If I’m wrong, Anthropic’s IPO will say so.
December 15, 2025 at 6:27 PM
Reposted by Jim Hawkins
edzitron.com
No, OpenAI will not make $13bn this year, and I believe it used Google Gemini and its "Code Red" as a means of creating hype around GPT-5.2, and to cover up ChatGPT's decreasing growth and popularity. The media hyped up Gemini 3 for no reason, too.
www.wheresyoured.at/ai-mythbuste...
“OpenAI Is Going To Make $13 Billion In 2025” Let’s start with some numbers: OpenAI paid Microsoft around $865.8 million through September 2025 as part of its 20% revenue share, from which we can extrapolate that OpenAI made $4.329 billion in that period. And this is accrual accounting, meaning those revenues and costs were booked in that period. Today I can report that for the same period, OpenAI was paid $198.14 million in royalties from Microsoft from sales of its AI models, bringing the revenue to the period to $4.527 billion. The idea that OpenAI “will make $13 billion in revenue” in this period is a farce, and that was the case before my reporting. The Information reported at the end of September that OpenAI made $4.3 billion in revenue through the first half of the year, and spent $2.5 billion on “the cost of revenue.” Assuming that were true (I do not believe it is), OpenAI would have to make more than double its first half revenue, and because OpenAI’s revenue is entirely subscription driven, I severely doubt that it can suddenly stimulate sales to that degree. Disney becoming a big ChatGPT subscriber is not going to save this company, nor will selling credits to generate more videos on Sora. ChatGPT, as a product, is effectively the same as it was at the beginning of 2025, apart from whatever it is the router model does on GPT-5. ChatGPT is as capable as it ever has been, has no new features that actually work, and ultimately lacks any real differentiation other than brand recognition. ChatGPT’s growth is stalling — flatlining in Europe according to Deutsche Bank. This company had to make $8 billion or more in the last six months, and I don’t think it has any path to meeting its end-of-year revenue goals. Gemini Was Not The Root Cause Of OpenAI’s Code Red Back in early December, Sam Altman called a “code red” in response to the launch of Google’s Gemini 3 (and its Nano Banana image generator), according to The Information: OpenAI CEO Sam Altman on Monday told employees he was declaring a “code red” to marshal more resources to improve ChatGPT as threats rise from Google and other artificial intelligence competitors, according to an internal memo. Sounds scary, Sam! What’s the plan to beat Google? Altman said Monday in an internal Slack memo that he was directing more employees to focus on improving features of ChatGPT, such as personalizing the chatbot for the more than 800 million people who use it weekly, including letting each of those people customize the way it interacts with them. Altman also said other key priorities covered by the code red included Imagegen, the image-generating AI that allows ChatGPT users to create anything from interior-design mockups to turning real-life photos into animated ones. Last month, Google released its own image generation model, Nano Banana Pro, to strong reviews. Altman said other priorities consisted of improving “model behavior” so that people prefer the AI models that powers ChatGPT more than models from competitors, including in public rankings such as LMArena; boosting ChatGPT’s speed and reliability; and minimizing overrefusals, a term that refers to when the chatbot refuses to answer a benign question. So let me get this straight, the plan is to make ChatGPT better, make image generation better, make people like the model more than other models, and make it better at benchmarks. Quick question, Sam: what the fuck have you been doing the last 11 months? Also, what is it that makes Gemini 3 better exactly? Nobody can seem to explain what it is. Sure, Gemini is better at benchmarks and Nano Banana is better at image generation. So? It’s another LLM, and does the same amount of things that other LLMs do. Quite literally a day after the Code Red story ran on The Information, they ran another story about OpenAI developing a model called ‘Garlic’ to counter Google: OpenAI, which in recent weeks has appeared to fall behind Google in AI development, is fighting back with a new large language model codenamed Garlic. Last week, OpenAI’s chief research officer Mark Chen told some colleagues about the new model, which was performing well on the company’s evaluations, at least when compared to Gemini 3 and Anthropic’s Opus 4.5 in tasks involving coding and reasoning, according to a person with knowledge of the remarks. ... But Garlic may be a bigger deal. Chen said OpenAI is looking to release a version of Garlic as soon as possible, which we think means people shouldn’t be surprised to see GPT-5.2 or GPT-5.5 release by early next year. OpenAI would launch GPT 5.2 just over a week later. There isn’t a chance in hell that OpenAI just started working on this in response to Gemini 3, and I’m guessing that while there’s a chance that OpenAI was truly rattled by Google, I think this may have been an attempt to get headlines for GPT 5.2, which would have otherwise been entirely ignored. Seriously, did you know that Claude Opus 4.5 launched? Did you even remember there was a Claude Opus 4.1? Hell, did anybody really talk about any previous Gemini announcements this much? I think there are a few potential scenarios: OpenAI is a fucking mess and rushed GPT 5.2 out the door because it got spooked by the press being nice about Gemini, and because normies liked Nano Banana Pro (and had already had Nano Banana out since August). OpenAI, desperately trying to ride Google’s coattails, created an artificial situation as a means of showing it was competitive, leading to a glossy Wall Street Journal story about how innovative ChatGPT is, a story from The Information about it “nearing 900 million weekly active users,” and Altman then told CNBC that there was “less of an impact then OpenAI feared.” I dunno, kinda feels like a manufactured situation! In any case, I think it’s worth considering that, at this point, Sam Altman just says shit whenever he feels like it.
December 15, 2025 at 6:27 PM
Reposted by Jim Hawkins
edzitron.com
All of those AI data centers are being built for imaginary demand. The only companies using lots of compute are giant companies building their own services - those who would need to rent GPUs either can't afford to or have their own.
www.wheresyoured.at/ai-mythbuste...
What This Means: The AI Data Center Market Will Have A Massive Oversupply Problem Once New Data Centers Are Built. There Is No Secret Wave Of Demand Putting aside that every AI company — be they a model or AI compute provider — is unprofitable, the current demand for AI services suggests that there is no secret wave of AI demand waiting in the wings. If there were, it would be impossible to get even 8 B200 GPUs from Lambda or Google Cloud or CoreWeave or Runpod or Nebius. A counterargument might be that there is demand, but that demand requires hundreds or thousands or tens of thousands of GPUs that don't currently exist on the market. If that were the case, there would be announcements of smaller, multi-million dollar compute deals, and indeed the commensurate whining from startups that desperately needed capacity and were unable to get it. While there might be blobs of millions of dollars of compute deals hiding inside those giant piles of RPOs, it seems impossible that those deals wouldn't somehow flow to a CoreWeave or Lambda or Nebius. In that case, this capacity revenue is already being booked, and it's fucking pathetic. We also wouldn't see so much price pressure, and we wouldn't be able to get $5.5-an-hour B200 GPUs from Nebius, or even be able to find one from Google Cloud, because if somebody is buying thousands or tens of thousands of them, there's not a chance in hell they're paying more than the sticker price. In simple terms, the companies using massive amounts of GPUs are hyperscalers forcing their AI services onto millions of people, OpenAI, Anthropic, and a handful of companies that run AI services that use their own models.
December 15, 2025 at 6:16 PM
Reposted by Jim Hawkins
edzitron.com
NVIDIA said that "GPUs are sold out in the cloud," and Nebius claims it is "sold out of all available capacity." Both are lying. You can rent Blackwells in minutes using Nebius, Lambda, or even Google cloud. There is no "insatiable demand."
www.wheresyoured.at/ai-mythbuste...
Sub-Myth: GPUs Are "Sold Out In The Cloud" So let's check 'em out! Nebius, another neocloud that grew out from Russian conglomerate Yandex, said in its Q2 2025 earnings release that its "annualized run-rate revenue guidance increased to $900 million to $1.1 billion for the end of 2025," meaning "$91 million a month," and proudly listed a bunch of stats about its revenue with quarter-over-quarter growth. It made $105 million in the quarter, up from $55 million the quarter before. A quarter later, and things didn't feel so exciting, with its Q3 2025 release feeling far more muted, failing to mention anything to do with revenue ($144 million) or growth other than that its expected 2026 annualized run-rate revenue would be "$7-9 billion," suggesting revenues would be, at best, $750 million for the year, which would — assuming Nebius makes a comparable amount in Q4 — only be about $300 million more in a year when it is supposedly adding 800 megawatts of capacity. Then there was this: Damn dude that's crazy! Anyway, here's me on Nebius' "us-central1" server, fully able to spin up 8 B200 Blackwell GPUs in a few clicks. AI demand is insatiable, bro! You simply cannot find these GPUs anywhere due to the insane demand that totally exists! But, uh, how about Lambda? I assume it’s sold out of Blackwells, right? That insatiable compute demand is making it impossible to find GPUs, I assume so. Anyway, time to drink this big cup of coffee and open my email... I don't even have anything funny to say! Jensen Huang said a few weeks ago that Cloud GPUs Are Sold Out, and he's just fuckin' lying! But Ed! Yes? Maybe he meant hyperscalers? It took me two minutes dicking around on Google Cloud to get to this screen which offered in a single click to deploy an instance with 8 B200 GPUs. I dunno, if Google has them available I'm going to guess Azure and Amazon Web Services do too (and both make it remarkably difficult to check without handing over money in advance). Google Cloud is a smaller player than both, and appears to be filthy rich with Blackwell GPUs. Then there's the thorny problem that no hyperscaler discloses AI revenue. We know (thanks to my reporting!) that Anthropic spent about $2.66 billion on AWS through the end of September, and OpenAI spent $8.67 billion on Azure in the same period, but who, otherwise, is spending oodles of cash on AI compute?
December 15, 2025 at 6:16 PM
Reposted by Jim Hawkins
edzitron.com
No, there is not "insatiable demand for AI compute." 75% of AI compute is bought by Microsoft, Meta, Amazon, OpenAI or NVIDIA - the company selling the GPUs. If demand were real, NVIDIA wouldn't have to buy $5-6bn of AI compute every year for the next 3 years.
www.wheresyoured.at/ai-mythbuste...
"There's Insatiable Demand For AI Compute" Did you know there's an insatiable demand for AI compute? Bain sure thinks so, and JP Morgan says that demand is "exploding, The Dispatch says that demand is booming," and Google's AI infrastructure head just said that it must double capacity every six months to meet AI demand. Yet when you look at the actual facts, things become a lot murkier. Let's look at the wonderful world of Neoclouds, and CoreWeave, my least favourite company in the world, a debt-ridden time bomb that I wish would explode already. CoreWeave's only business line is renting out GPUs for AI compute, which would mean, in theory, that its revenues would be going absolutely batshit right now as companies bang on the door, screaming "I must run AI models, give me access to the latest GPUs!" CoreWeave's earnings do not paint a picture of insatiable demand. In its most recent quarter, CoreWeave made $1.36 billion in revenue (and lost $110 million), and claims to have a $55.6 billion backlog with capacity issues caused by "a third-party data center developer that's behind schedule," referring to Core Scientific, before claiming that the delay wouldn't affect CoreWeave's backlog, which it absolutely will. Here's the thing: CoreWeave constantly advertises GPU availability, including the latest-generation Blackwell GPUs, with instances available of B200 GPUs and GB200 pods available right now. Nevertheless, let's go back to those earnings. One customer accounted for 67% of CoreWeave's revenue, with four customers — A, C, D and E — responsible for 35%, 31%, 10% and 10% of its accounts receivable, meaning that four customers account for 86% of CoreWeave's outstanding revenue, or $1.42 billion. On average, CoreWeave also takes about 94 days to get paid (per its Day Sales Outstanding) which is really bad for a company that loses tons of money. It’s also worse than three-quarters of the software industry, and genuinely concerning when you consider that CoreWeave is in an industry with high capital (data centers and GPUs) and operational (especially power) expenses, and it’s burdened with a formidable debt load that absolutely must be serviced, in part because it’s using its GPUs as collateral. In any case, the majority of CoreWeave's revenue comes from a few big clients — OpenAI ($22.4 billion), Microsoft (for OpenAI), Google (for OpenAI), Meta ($14 billion for...something?) and, of course, NVIDIA, which, as I mentioned previously, signed a deal to buy $6.3 billion of unused capacity from CoreWeave through 2032, expanding its original $1.4 billion deal. CoreWeave’s customer case studies include Mistral, IBM (whose CEO just said there was "no way" spending trillions on AI data centers would pay off), somebody called "Neuralift AI," a company that has raised a total of $3 million in funding, a company called "Moonvalley" (total funding: $154.1M), Authentik (no funding raised), Cerebrium ($8.5 million raised that CoreWeave adds is in a partnership with Decart, which has raised a little under $200 million), Perception AI (leading to this video from two years ago), Product Insight (leading to a two-year-old article about "GPU-accelerated medical animations), Rad.ai (which has raised over $140 million, with its most recent $60m round from January), Trillion Labs (which raised a total of $4.2 million), UbiOps ($2.2 million in funding) and vCluster ($24 million in funding). Hedge fund Jane Street is also present, and CoreWeave claims to be building some sort of "flexible, full stack solution" to train its "next generation of training models." OpenAI is also there, I guess. I listed these companies out to make a point: that the majority of CoreWeave's "case studies" are for companies that are going to provide (or be able to afford) tens or hundreds of thousands of dollars of business — millions, at best — and t…
December 15, 2025 at 6:16 PM
Reposted by Jim Hawkins
deanobeidallah.bsky.social
Trump's vile response to Rob and Michele Reiner's murder is simply a reminder of how evil Trump is. And yes the word is EVIL.
December 15, 2025 at 6:27 PM
Reposted by Jim Hawkins
sundersays.bsky.social
3 examples of this "time and again" scenario (that sounds imagined by Lord Sewell)?
December 15, 2025 at 6:27 PM
jimallthetime.bsky.social
I need to show you these figures because I’m very excited and hugely grateful to you for making this happen 🍾

Here’s how Hello You the Shropshire podcast is performing on Spotify … 🎧

Thank you so much for your support 🙏🏻

Hear more in Hello You Episode 149a 😁

open.substack.com/pub/jimallth...
December 15, 2025 at 5:52 PM
jimallthetime.bsky.social
I need to show you these figures because I’m very excited and hugely grateful to you for making this happen 🍾

Here’s how Hello You the Shropshire podcast is performing on Spotify … 🎧

Thank you so much for your support 🙏🏻

Hear more in Hello You Episode 149a 😁

open.substack.com/pub/jimallth...
December 15, 2025 at 5:51 PM
Reposted by Jim Hawkins
charlotte2153.bsky.social
No idea what telly Lord Sewell is watching, because I can't think of a single show with this plotline let alone it being something that is inescapable for young white men who subsequently now face a crisis of masculinity drawing them to the far right?? 🙃
Screenshot from The Times article 'Theres a Reason Young Men Turn to Nick Fuentes' reading "Watching contemporary TV drama often requires the suspension of disbelief. Not simply because plots are implausible, but because the moral universe feels inverted. Time and again, we are presented with the same tableau: a five-foot-five female heroine effortlessly subdues a 20-stone man, pins him by the neck, denounces him as a “racist, fascist, misogynist” and dispatches him with moral and literal finality. Masculinity is caricatured as brute, stupid and disposable. Power is no longer something to be questioned; it is something to be reassigned"
December 15, 2025 at 1:52 PM
Reposted by Jim Hawkins
tpgroberts.bsky.social
Feel like this also misses the point that young men aren't really being drawn to Nick Fuentes in any significant numbers
charlotte2153.bsky.social Charlotte Nichols MP @charlotte2153.bsky.social · 14h
No idea what telly Lord Sewell is watching, because I can't think of a single show with this plotline let alone it being something that is inescapable for young white men who subsequently now face a crisis of masculinity drawing them to the far right?? 🙃
Screenshot from The Times article 'Theres a Reason Young Men Turn to Nick Fuentes' reading "Watching contemporary TV drama often requires the suspension of disbelief. Not simply because plots are implausible, but because the moral universe feels inverted. Time and again, we are presented with the same tableau: a five-foot-five female heroine effortlessly subdues a 20-stone man, pins him by the neck, denounces him as a “racist, fascist, misogynist” and dispatches him with moral and literal finality. Masculinity is caricatured as brute, stupid and disposable. Power is no longer something to be questioned; it is something to be reassigned"
December 15, 2025 at 2:11 PM
jimallthetime.bsky.social
This!

substack.com/@tiphainemar...
Tiphaine Marie (@tiphainemarie)
A little reminder as the days are getting colder
substack.com
December 15, 2025 at 2:18 PM
Reposted by Jim Hawkins
garybutler.bsky.social
"Some people are so poor, all they have is money" - Bob Marley
December 15, 2025 at 9:41 AM
Reposted by Jim Hawkins
komaniecki.bsky.social
They cannot understand art, or community, or even real joy! Their souls have been devoured by their own greed
December 15, 2025 at 8:20 AM
Reposted by Jim Hawkins
komaniecki.bsky.social
For both Cuban and Musk, my emotions waver between revulsion and pity. They have all material things a person could want. They influence policy, culture, and how information is distributed. But their wealth literally precludes them from understanding the human condition. I think I would rather die
December 15, 2025 at 8:18 AM
Reposted by Jim Hawkins
komaniecki.bsky.social
I have grown to believe that excessive wealth does something to your brain that is analogous to a serious head injury
December 15, 2025 at 6:03 AM
Reposted by Jim Hawkins
socialistdogmom.bsky.social
i don’t know why tech guys think they are creative storytellers or whatever but that is actually what my job is and i will never ask the robot for help. the painful process of working out how to tell the story is not a bug. that’s not a problem in need of solving.
December 15, 2025 at 2:17 PM
Reposted by Jim Hawkins
socialistdogmom.bsky.social
what if we lock these guys inside of a
museum until they develop some respect for art or the statues come alive at night and eat them. either way, just get this shit out of my face.
komaniecki.bsky.social Robert Komaniecki @komaniecki.bsky.social · 22h
I have grown to believe that excessive wealth does something to your brain that is analogous to a serious head injury
December 15, 2025 at 2:13 PM
jimallthetime.bsky.social
@radio.today
jimallthetime.bsky.social Jim Hawkins @jimallthetime.bsky.social · 1d
Every single word of this is true, heartfelt and powerful. *This* is why I grieve for the radio we've lost, the radio we used to have. I'm proud to have worked in radio when radio mattered

Please read this closely. I feel it deeply

Losing local radio — and my mum

www.livpost.co.uk/losing-local...
Losing local radio — and my mum
Merseyside stations soundtracked David Lloyd’s relationship with his mother. Now he’s grieving a double loss
www.livpost.co.uk
December 15, 2025 at 12:33 PM
Reposted by Jim Hawkins
bencollins.bsky.social
Elon's little serfs are quite literally begging journalists to come back because every publicly available source of data shows people have fled the platform and young people won't use it.

They are literally begging regular people to come back. Don't do it! It is a website by and for actual Nazis!
An interview with Twitter's Chief Product Officer title "X Wants Its Haters Back"
December 15, 2025 at 1:27 AM
jimallthetime.bsky.social
Trump’s takeover of the Kennedy Center should serve as a warning to UK arts institutions

www.theguardian.com/commentisfre...
Trump’s takeover of the Kennedy Center should serve as a warning to UK arts institutions | Charlotte Higgins
Art is not immune from the advances of the hard right – but it isn’t too late for resistance, says the Guardian’s chief culture writer, Charlotte Higgins
www.theguardian.com
December 15, 2025 at 12:01 PM
Reposted by Jim Hawkins
jimallthetime.bsky.social
Every single word of this is true, heartfelt and powerful. *This* is why I grieve for the radio we've lost, the radio we used to have. I'm proud to have worked in radio when radio mattered

Please read this closely. I feel it deeply

Losing local radio — and my mum

www.livpost.co.uk/losing-local...
Losing local radio — and my mum
Merseyside stations soundtracked David Lloyd’s relationship with his mother. Now he’s grieving a double loss
www.livpost.co.uk
December 14, 2025 at 8:47 AM
Reposted by Jim Hawkins
jimallthetime.bsky.social
Pictures from Christmas at Attingham Park
December 13, 2025 at 9:17 PM
Reposted by Jim Hawkins
adamjschwarz.bsky.social
Ahmed al Ahmed, a Lebanese Muslim who owns a Sydney fruit shop, risked his life to disarm one of the Bondi Beach shooters.

His courage gives me hope amidst the horror and darkness. Muslims and Jews are not enemies. Our futures, in the Middle East and beyond, are intrinsically bound.

A true hero.
December 14, 2025 at 1:41 PM