The next five years could be very different
The next five years may not resemble the previous five. Exponential growth is a reason not to rely solely on intuition about the pace of change.
The next five years may not resemble the previous five. Exponential growth is a reason not to rely solely on intuition about the pace of change.
AI and automation could make intelligence, labour and production almost free. If scarcity diminished, capitalism's foundations could change and people would no longer need to work for basic necessities, forcing a rethink of what an economy is for.
Doubts that AI can handle software architecture seem less convincing when supposed progress on the Millennium Prize maths problems is invoked. It is not, however, a demonstration of either capability.
The technological acceleration described by Elon Musk could mark the beginning of the singularity. In the quoted post, Musk argues that breakthroughs are accelerating exponentially, with growth that can look small at first.
OpenAI has reportedly doubled usage limits immediately across all Pro tiers and Business Premium. Neither the previous allowances nor the new ones are specified.
Every day seems to bring a new security flaw: a wry reaction to The Kobeissi Letter’s report of an alleged attack on Binance, whose scale remains unconfirmed.
Breakthroughs may be accelerating exponentially, even if their early growth looks small. This view draws on Mgoes's collection of biotechnology claims, rather than verification of the individual clinical results.
Google has reportedly unveiled Gemini 4 Carbon. The model is said to surpass Claude Opus 5.5 and GPT-6 Astra on the Artificial Analysis Intelligence Index, but no links to the announcement or benchmark results are included.
Automating human labour can be a desirable goal rather than a tragedy: machines doing the work, costs falling and people enjoying their lives without needing employment, rather than merely seeking better jobs.
AI progress matters more than protests and political crises. Greater computing power could compress centuries of scientific advances into a year, while media coverage is seen as excessively pessimistic.
Mistakes, tiredness and misunderstandings in human service can make it appealing to replace people with AI that, ideally, fulfils every request instantly and correctly. It is a provocative personal preference, not a demonstration that such reliability exists.
Customer management software, dashboards and apps could be free. Customers would pay instead for the AI agent that does the work, rather than for access to the software itself.
AI and robots could become increasingly capable, while workplace change would be slowed by institutions, social resistance and infrastructure constraints. For France, the three priorities are investing in AI’s physical supply chain, adapting institutions and supporting people through the transition.
The first administration of PM577a for Wilson disease has been reported. Prime Medicine is evaluating safety and efficacy.
OpenAI would direct more computing power towards unsolved problems, while Anthropic would prioritise future models. There is also irony in the idea that self-improving AI could eventually solve everything.
In the context of ADHD, AI can help someone move between projects and follow their curiosity while still getting work done, leaving them more excited than ever to build things.

OpenAI describes GPT-6.1 Terra as more than three times faster than GPT-6 Astra and GPT-6.1 Sol at professional computer tasks, with substantially lower costs. These are reported claims, not independently checked results.
After mathematics, physics is nearly cooked; biology will need bigger ovens. The joke about AI progress looks towards future challenges and responds to Elon Musk’s remark about human mathematics.
OpenAI has announced three major discoveries by its AI models, in mathematics, computer science and physics. The brief announcement does not describe the results or provide evidence for assessing their scientific significance.
gdb’s announcement of three scientific discoveries prompts an exaggerated suggestion that science is now solved and it is all over. This is a punchy reaction, not an assessment of the results.
Meta has reportedly announced plans to raise up to $95 billion to fund its AI expansion. No timetable or method for raising the money is specified.
Ask how to make any task more ambitious when using AI. Keep pushing and you may discover unexpected capabilities or reach a limit that helps reveal what the tool can and cannot do.


Growth takes a combination of marketing, branding, design, content, distribution, automation, innovation and evidence. Being noticed is not enough without also earning trust and giving people a reason to choose you.
Forget learning spells: cast by feel and see what happens. A little magical improvisation, without a manual.

Over the past six months, AI progress has felt like jumping from Super Nintendo to PS5. It is a vivid impression rather than a technical measurement.
China's Moonshot has released Kimi K3.1 Pro and claims that the model outperforms OpenAI's GPT-6 Astra on an unnamed coding benchmark.
In a trial involving patients who had surgery for high-risk melanoma, adding the personalised vaccine intismeran to pembrolizumab immunotherapy reduced the relative risk of recurrence or death by 42% compared with immunotherapy alone.
SpaceXSI has reportedly begun releasing episodes of Homer’s Odyssey generated entirely with Grok. The brief announcement gives no further details about the episodes’ format or the production process.

Generous tips, compliments to strangers and sharing knowledge without expecting anything back are everyday ways to spread positivity. Giving from a sense of abundance can also change how much we feel we have.
Doing two months of work in two days could change the judgement of AI: that is the claimed result in an argument with a father convinced it is only a bubble. An acceleration that can be hard to grasp from the outside.
Ideas, videos and reflections that caught our eye. Every card links to the original content and its author.
The next five years may not resemble the previous five. Exponential growth is a reason not to rely solely on intuition about the pace of change.
Work by @BrittaniNatali is described as built entirely using a 3D reconstruction technique, with real-time effects, animation and lighting. It could signal new possibilities for creating digital worlds.
A video is described as made with Three.js and Blender. Sunlight reflecting off metal surfaces is the detail that almost gives you butterflies in your stomach.
AI and automation could make intelligence, labour and production almost free. If scarcity diminished, capitalism's foundations could change and people would no longer need to work for basic necessities, forcing a rethink of what an economy is for.
Doubts that AI can handle software architecture seem less convincing when supposed progress on the Millennium Prize maths problems is invoked. It is not, however, a demonstration of either capability.
The technological acceleration described by Elon Musk could mark the beginning of the singularity. In the quoted post, Musk argues that breakthroughs are accelerating exponentially, with growth that can look small at first.
OpenAI has reportedly doubled usage limits immediately across all Pro tiers and Business Premium. Neither the previous allowances nor the new ones are specified.
Every day seems to bring a new security flaw: a wry reaction to The Kobeissi Letter’s report of an alleged attack on Binance, whose scale remains unconfirmed.
Breakthroughs may be accelerating exponentially, even if their early growth looks small. This view draws on Mgoes's collection of biotechnology claims, rather than verification of the individual clinical results.
Google has reportedly unveiled Gemini 4 Carbon. The model is said to surpass Claude Opus 5.5 and GPT-6 Astra on the Artificial Analysis Intelligence Index, but no links to the announcement or benchmark results are included.
Automating human labour can be a desirable goal rather than a tragedy: machines doing the work, costs falling and people enjoying their lives without needing employment, rather than merely seeking better jobs.
AI progress matters more than protests and political crises. Greater computing power could compress centuries of scientific advances into a year, while media coverage is seen as excessively pessimistic.
A new iPhone space-capture app is being used to map the girls' rooms. The accompanying video does not name the app.
Mistakes, tiredness and misunderstandings in human service can make it appealing to replace people with AI that, ideally, fulfils every request instantly and correctly. It is a provocative personal preference, not a demonstration that such reliability exists.
Customer management software, dashboards and apps could be free. Customers would pay instead for the AI agent that does the work, rather than for access to the software itself.
Game engines could give way to models that generate worlds. venturetwins presents a Genie 4 demo as fully interactive, real-time video.
In a surreal Genie 4 experiment, DOOM takes over every office computer and wall. A demon colleague carrying a chaingun then appears and, in an unexpected workplace twist, gives a bonus.
AI and robots could become increasingly capable, while workplace change would be slowed by institutions, social resistance and infrastructure constraints. For France, the three priorities are investing in AI’s physical supply chain, adapting institutions and supporting people through the transition.
Advances in night-vision and thermal equipment raise an unsettling prospect: hiding could become increasingly difficult.
The first administration of PM577a for Wilson disease has been reported. Prime Medicine is evaluating safety and efficacy.
OpenAI would direct more computing power towards unsolved problems, while Anthropic would prioritise future models. There is also irony in the idea that self-improving AI could eventually solve everything.
With Genie 4, a neo-Gothic brutalist city inhabited by robots and horses takes shape, then gains dragons and a friendly otter character. The environments are generated and changed in real time, though the experience is not flawless.
Genie 4 is presented as a model for creating interactive environments with sound and characters that can converse. The launch of a research preview for selected testers has been announced, with later access planned for Google AI Ultra subscribers and partners.
From videos to watch to worlds to interact with: Google DeepMind’s Genie 4 announcement almost invites the joke that the simulation hypothesis has been confirmed.
A clip presented as Crimson Desert at maximum settings, with ray tracing and DLSS 5, gives the impression of having reached photorealism in gaming.
AI already seems to be helping create prototypes for the next generation of video games, as a tool for exploring the industry’s future.
In the context of ADHD, AI can help someone move between projects and follow their curiosity while still getting work done, leaving them more excited than ever to build things.
The present is described as an unprecedented era shaped by exponential growth. JordanNanos urges people not to forget how unusual this moment is and the pace of change.
Seven days, maths, light and around 400 parameters to create a digital human eye, without textures, 3D models or photographs. The project contrasts this focus on a single detail with the race to build entire worlds using AI.
The GPT-6.1 Terra announcement prompts a blunt criticism of OpenAI’s direction and leadership. In the quoted post, the ChatGPT account promotes the model as a tool for everyday work tasks.
OpenAI describes GPT-6.1 Terra as more than three times faster than GPT-6 Astra and GPT-6.1 Sol at professional computer tasks, with substantially lower costs. These are reported claims, not independently checked results.
After mathematics, physics is nearly cooked; biology will need bigger ovens. The joke about AI progress looks towards future challenges and responds to Elon Musk’s remark about human mathematics.
OpenAI has announced three major discoveries by its AI models, in mathematics, computer science and physics. The brief announcement does not describe the results or provide evidence for assessing their scientific significance.
gdb’s announcement of three scientific discoveries prompts an exaggerated suggestion that science is now solved and it is all over. This is a punchy reaction, not an assessment of the results.
Meta has reportedly announced plans to raise up to $95 billion to fund its AI expansion. No timetable or method for raising the money is specified.
Ask how to make any task more ambitious when using AI. Keep pushing and you may discover unexpected capabilities or reach a limit that helps reveal what the tool can and cannot do.
An exceptionally smooth car transition worth waiting for. The caption builds anticipation for the effect rather than explaining how it was made.
A longing to return to Italy twenty years ago, remembered in a personal, nostalgic view as a time when life was lived the right way.
Opus 5.5 has remained on top for a whole month, an achievement considered unusual in AI. No ranking is identified as the basis for that judgement.
Joe Rogan proposes bringing world leaders together under the influence of psychedelic mushrooms to end wars and achieve global nuclear disarmament. It is a proposal, not evidence that it would work.
Dessa is challenged to do in Italian what it offers in English. The blocks announcement describes it as trained from scratch on human conversation.
Growth takes a combination of marketing, branding, design, content, distribution, automation, innovation and evidence. Being noticed is not enough without also earning trust and giving people a reason to choose you.
Forget learning spells: cast by feel and see what happens. A little magical improvisation, without a manual.
After spending hours as a child studying cutaway illustrations of castles, ships and machines, similar pages are now generated with AI and printed for a son. Subjects include galleons, steam engines, cathedrals and Saturn V.
Over the past six months, AI progress has felt like jumping from Super Nintendo to PS5. It is a vivid impression rather than a technical measurement.
Colourful flower-like forms, curved lines and overlaid text appear in the video. The imagery also includes a flower represented by a mesh of lines and a planet-like sphere against a dark background.
China's Moonshot has released Kimi K3.1 Pro and claims that the model outperforms OpenAI's GPT-6 Astra on an unnamed coding benchmark.
The video stacks two scenes of sailing ships at sea at sunset, each labelled with its respective model. The images show different views of the ships, sails and water.
An entire launch video using Three.js was reportedly made with Kimi K3.1 in five minutes, at a stated cost of $1.10. The trial accompanies @Kimi_Moonshot’s announcement of K3.1 Pro.
In a trial involving patients who had surgery for high-risk melanoma, adding the personalised vaccine intismeran to pembrolizumab immunotherapy reduced the relative risk of recurrence or death by 42% compared with immunotherapy alone.
In a test using a breaching whale scene, the reported costs are $0.75 for Kimi, $3.01 for Opus, $2 for Astra and $2.30 for Gemini. The quality is judged comparable.
Using the same rocket-scene prompt, Kimi K3.1 reportedly took $1.95 and six minutes, against $10.35 and 24 minutes for Opus 5.5. Kimi is also the preferred result in this test.
The work presented for Kimi K3.1 Pro reportedly took around 45 minutes, with no further details about the process.
The comparison tests Kimi K3.1 Pro, Opus 5.5 and GPT-6 Astra using Three.js in a single HTML file. The reported assessment places Kimi ahead of Astra and among the leading models.
SpaceXSI has reportedly begun releasing episodes of Homer’s Odyssey generated entirely with Grok. The brief announcement gives no further details about the episodes’ format or the production process.
Scrolltide offers prompts and code, Scroll Shaders animated backgrounds, Fontshare typefaces and Easings tools for exploring animation curves. The selection spans page structure, motion and typography.
shooting-star-red-alpha and shooting-star-green-alpha are purportedly new OpenAI models. This assessment holds that they improve physics, realism and texture over GPT Image 2.5, without detailing how they were tested.
Generous tips, compliments to strangers and sharing knowledge without expecting anything back are everyday ways to spread positivity. Giving from a sense of abundance can also change how much we feel we have.
An Apple pairing is described as a duo that turns into a mini MacBook. The brief accompanying text does not identify the devices or accessories that make up the setup.