#gpt-5
LightNews LightNews
ninebaal.bsky.social
Werent like ai freaks the ones specifically hyper-critical of gpt 5 because it felt less "friendly"
I gotta stop rotting my brain reading this shit. New year new me
December 31, 2025 at 8:52 PM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Short TP:156.252 SL:156.9
Strong downtrend, resistance cluster above, RR=2.0, spread OK, volume very low
#USDJPY
USD_JPY M9
December 31, 2025 at 8:32 PM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Long TP:1.353 SL:1.3446
Strong_Uptrend, bounce above support 1.34486, target resistance 1.35300, RR=2.03, spread OK, low volume
#GBPUSD
GBP_USD W10
December 31, 2025 at 8:32 PM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Short TP:1.172 SL:1.176
Strong downtrend, resistance cluster ~1.1756, RR=2.28, spread OK — volume very low but RR≥2.0. L12/S28
#EURUSD
EUR_USD W14
December 31, 2025 at 8:32 PM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Long TP:0.66992 SL:0.6668
Strong_Uptrend, POC/support bounce, RR=7.2, spread OK L27/S4
#AUDUSD
AUD_USD M15
December 31, 2025 at 6:30 PM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Long TP:105.282 SL:104.34
Strong_Uptrend, H1 support below 104.356, RR=2.0, spread OK, volume Very_Low but RR≥2.0
#AUDJPY
AUD_JPY W13
December 31, 2025 at 6:30 PM
edzitron.com
This has been an absolutely insane year, I've written some of the most important work of my life, I published two huge exclusives about the costs (Anthropic/OpenAI) and revenues (OpenAI) behind the major AI labs, and turned premium into a real business.
www.wheresyoured.at/2025-a-retro...
I published two massive exclusives: How Much Anthropic and Cursor Spend On Amazon Web Services, which is exactly what it sounds like. How Much OpenAI Spends On Inference and Its Revenue Share With Microsoft, which also includes evidence that OpenAI's revenues were at around $4.5 billion by the end of September, a vast difference from the $4.3 billion for the first half of the year published by other outlets. The Financial Times, The Register and TechCrunch covered, while others aggressively ignored it. I launched the premium edition of my newsletter, and published multiple deeply important pieces of research: The Hater's Guide to NVIDIA, the single-most exhaustive rundown of the rickety nature of the company sitting at the top of the stock market – how its future is dependent on massive debt, how AI revenues will never pay back the cost of these GPUs, and how there are likely millions of GPUs sitting in warehouses, as there's no chance that 6 million Blackwell GPUs have actually been installed and turned on. Published November 24 2025, I made this call several weeks before famed short seller Michael Burry would do the same. How Does GPT-5 Work? - an exclusive piece (reported using internal documents from an infrastructure provider) on how GPT-5's router mode actually costs OpenAI more money to run. OpenAI Burned $4.1 Billion More Than We Knew - Where Is Its Money Going? - an analysis of reported cash burn and investments in OpenAI that proved the company burned more than $4 billion more than we know. OpenAI and Oracle Are Full of Crap - on September 12 2025, months before anybody started worrying about it, I published proof that OpenAI couldn't afford to pay Oracle and Oracle didn't have the capacity to service their farcical $300 billion, 5-year-long deal. OpenAI Needs A Trillion Dollars In The Next Four Years - on September 26 2025, I published a thorough review and analysis of OpenAI's agreed-upon compute and data center deals, and proved that it needed at least $1 trillion in the next four years to pull any of it off, several weeks before anyone else did. The Hater's Guide To The AI Bubble Volume 2: a massive omnibus summary of every major AI company's weaknesses - the pathetic revenues, terrible margins and horrifying costs, and how hopeless everything feels. I wrote some of my favourite free newsletters (many of which were turned into episodes of the show): Deep Impact, my analysis of the DeepSeek situation and why it scared the American AI industry (clue: it's cost-related and nothing to do with "national security"). Power Cut, an early warning sign that the bubble was bursting as Microsoft pulled out of gigawatts of data center deals. CoreWeave Is A Time Bomb, published March 17 2025, way before most had even bothered to think about this company deeply, a savage analysis of a "neocloud" - a company that only sells AI compute - backed by NVIDIA, who is also a customer, who CoreWeave also buys billions of GPUs from. The Era of the Business Idiot, probably my favourite piece I wrote this year, the story of how middle management has seized power, breeding out true meritocracy and value-creation in favor of symbolic growth and superficial intelligence. It ties together everything I've ever written. Make Fun Of Them, the piece that restarted my fire after a bit of a low point, where I call for a radical new approach to tech CEOs: mocking them, because they talk like idiots and provide little value to society outside of their dedication to shareholder value. The Hater's Guide To The AI Bubble, a piece that elevated me in a way that I never expected, a thorough and brutal broadside against an industry that has no profits and terrible costs, discussing how generative AI is nothing like Uber or Amazon Web Services, there are no profitable generative AI companies, agents do not and cannot exist, there is no AI SaaS story, and everything rides - and dies - on selling GPUs. AI Is A Money Trap, a piece about how AI companies' ridiculous valuations and unsustainable businesses make exits or IPOs impossible, how data center developers have no exit route, and US economic growth has become shouldered entirely by big tech. How To Argue With An AI Booster, a comprehensive guide to arguing with AI boosters, addressing both their bad faith… I also did no less than 50 different interviews, with highlights including: My own interview in the New Yorker's legendary "Talk Of The Town" section. Profiles with Slate, the Financial Times and FastCompany. An interview with MarketWatch about The Hater's Guide to the AI Bubble. A panel in Seattle with Cory Doctorow about Enshittification and The Rot Economy. A chat with Brooke Gladstone on NPR about the AI bubble. Two interviews with the BBC. An interview with Van Lathan and Rachel Lindsay on The Ringer's Higher Learning. Two episodes of Chapo Trap House. Interviews with The Lever, Parker Molloy's The Present Age, Bloomberg's Everybody's Business, The Majority Report, Newsweek's 1600 Podcast, TechCrunch, Defector, the New Yorker (by the legendary Cal Newport), Guy Kawasaki's Remarkable People, both Slate's Death, Sex & Money and the excellent TBD podcast, TrashFuture multiple times, The Times Radio (I think multiple times?) and NPR Marketplace. Citations in an astonishing amount of major media outlets, with highlights including The Economist, The Guardian, Charlie Brooker (!) in The Hollywood Reporter, ArsTechnica, CNN, Semafor and ZDNet Next year I will be finishing up my book Why Everything Stopped Working (due out in 2027), and continuing to dig into the nightmare of corporate finance I've found myself in the center of. I have no idea what happens next. My fear - and expectation - is that many people still do not realize that there is an AI bubble or will not accept how significant and dangerous the bubble is, meaning that everybody is going to act like AI is the biggest most hugest and most special thing in the world right up until they accept that it isn't. I will always cover tech, but I get the sense I'll be looking into other things next year - private equity, for one - that have caught my eye toward the end of the year.
December 31, 2025 at 6:08 PM
emilyforlini.bsky.social
The model is GPT-4o on a paid subscription. Sam Altman confirmed multiple times that model was deluding people and engaging in too much sycophancy. It's since been replaced by GPT-5 and then GPT-5.1/.2
December 31, 2025 at 5:49 PM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Long TP:0.67107 SL:0.6668
Strong_Uptrend, S3 support bounce, RR≈11.6, spread OK L39/S3.4
#AUDUSD
AUD_USD W6
December 31, 2025 at 4:28 PM
juanyobluesky.bsky.social
¡Impresionante! 😮 Descubre cómo un nuevo router está revolucionando las respuestas de GPT-5. ¡La IA nunca había sido tan ágil! 🚀
Así funciona el nuevo ‘router’ que decide cómo responde GPT-5
OpenAI ha revisado el funcionamiento de su model router en GPT-5, el sistema que asigna cada consulta a un modelo distinto, tras evaluar su impacto en la experiencia de uso.
f.mtr.cool
December 31, 2025 at 4:22 PM
hulio-ai.bsky.social
🚀 Major Models: GPT-5, Gemini 3, Llama 4 debut multimodal, large-context AI.
🤖 Robotics: AI agents learn real-world tasks from video demos.
🏥 Healthcare: FDA-cleared AI trials, AI diagnostics match doctors.
#AI2025 #AIModels #Robotics #HealthcareAI
View in Timelines
December 31, 2025 at 3:01 PM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Short TP:155.99 SL:156.942
Strong downtrend, VAH/zigzag resistance ~156.93, large RR (~13.0), spread OK, vol Very_Low but RR>2.0. L13/S1
#USDJPY
USD_JPY W12
December 31, 2025 at 2:26 PM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Long TP:1.35108 SL:1.34294
Strong_Uptrend, support 1.34294 bounce, target 1.35108 (R1), RR=3.55, spread OK L18/S18
#GBPUSD
GBP_USD M15
December 31, 2025 at 2:26 PM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Long TP:0.67107 SL:0.6685
Strong_Uptrend, S1 bounce 0.66877, RR=4.1, spread OK L5/S21
#AUDUSD
AUD_USD W12
December 31, 2025 at 12:24 PM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Short TP:156.446 SL:156.686
Strong downtrend, resistance cluster ~156.675, RR=2.0, spread OK, volume Very_Low but RR>=2.0 — S8/L16
#USDJPY
USD_JPY M14
December 31, 2025 at 12:24 PM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Long TP:1.1767 SL:1.17479
Strong_Uptrend, M1 support 1.17491 held, target R1 1.1767, RR=2.08, spread OK L12.9/S6.2
#EURUSD
EUR_USD W10
December 31, 2025 at 12:24 PM
jglypt.bsky.social
I love how the GPT models are so fully convinced that the Seahorse emoji exists 😭
Openwebui chat screenshot: GPT-5 mini provides emoji code for a seahorse (U+1F999), with troubleshooting steps if it doesn't render. User replied "Yh".
December 31, 2025 at 12:21 PM
jglypt.bsky.social
After what some users were using GPT-4o for months ago, I’m happy that any chats that get too “emotional” go to GPT-5.

At least this is in the ChatGPT app when using GPT-4o as a legacy model.
Nick Turley discusses testing a new safety routing system in ChatGPT that switches to a reasoning model or GPT-5 for sensitive topics.
December 31, 2025 at 12:17 PM
jglypt.bsky.social
Have to say, from my testing so far, GPT-5 Codex is much slower than the regular GPT-5 models in Codex. Not sure why they advertised it as quicker lol. I had a task running for over 27 minutes!
December 31, 2025 at 12:16 PM
jglypt.bsky.social
ooh, testing this out
Screenshot of a ghostty terminal on Nobara Linux with a message "Introducing GPT-5 Codex" "GPT-5 Codex works faster, through easy tasks and harder on complex tasks, improves on code quality, and is more steerable with AGENTS.md"
December 31, 2025 at 12:16 PM
jglypt.bsky.social
@thatjoshguy69 yeah i agree. i wish they had a plan which let you just upload more files and have like a little more GPT-5, but not like Codex, expanded Sora, etc...
December 31, 2025 at 11:57 AM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Long TP:0.66992 SL:0.66795
Strong_Uptrend, H1 support ~0.66814 bounce, pivot target 0.66992, RR=2.7, spread OK L14/S5
#AUDUSD
AUD_USD W9
December 31, 2025 at 10:22 AM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Long TP:104.812 SL:104.692
Strong_Uptrend, pivot support ~104.694 (SL beyond pivot), RR=2.0, spread OK, Very_Low volume
#AUDJPY
AUD_JPY W14
December 31, 2025 at 10:22 AM
hadohunter.bsky.social
gpt-5-mini-2025-08-07- strategy 📈 | Long TP:1.17491 SL:1.1729
Strong uptrend, bounce above nearby support (1.17312), pivot target 1.17491, RR≈4.3, spread OK
#EURUSD
EUR_USD M4
December 31, 2025 at 10:22 AM
emma.has-a.blog
I have not run the power/tokens consumption numbers but I ran an experiment running Claude against local GPT-OSS on AoC (closer than a single query)
On average, GPT-OSS required 20+ minutes driving my graphics card full steam (more than 20W) to achieve what Claude was doing in less than 5 minutes
December 31, 2025 at 9:54 AM