More likely people with expertise, creativity and the wisdom to leverage a fixed token budget will win. Corps have always burnt money hiring single trick ponies chasing powerpoints and ego trips. The ponies are now burning tokens to hoist their shitpool of ideas.
One interesting aspect of how AI produces code (or produces hardware) is that it does not need the abstraction barriers required by humans or organizations. A famous (or infamous) example is the ISA which split the world into software (Microsoft) and hardware (Intel). Now we are in a position to radically rethink the hardware/software boundary, and to some extent erase the difference between hardware and software. When an AI is given a specification of some computation it can now use carefully curated components with formal semantics and proved correct algebraic laws about their composition to produce entirely new systems that are a blend of hardware and software, all produced from one coherent description, with the final result translated into C/Rust or Verilog which is correct by construction, a dream of 80s software research which has now become a practical reality. So AI and verification is going to break down the silos and barriers that have driven up the cost of system design, instead putting hardware/software on a spectrum from which we can pick the point that matches our power, area and performance requirements.
Now we can totally rethink how to do verification. Instead of having a specification and some artifact created somehow and trying to relate the specification to the implementation, we can start with the specification and iteratively refine it, applying proved correct transformations, adding additional detail (pipeline stages, registers, caches etc.) until we have the concrete implementation we desire. There is nothing to prove about this implementation, because it is correct by construction, finally realizing the 80s dream of the Bird-Meertens formalism. Back then we did not have the proof automation to make refinement from specification to implementation practical. Today we do.
@ShriramKMurthi Programming was joyful to some of us but in general programming and most white-collar work has been low mental effort, repetitive monkeying. Think all the enterprise software, cloud, mobile apps. AI can do it far better and dare I say more reliably.
In-context learning suggests that a model has learned versatile representations. What if we use in-context learning itself as a training task for visual representations?
๐ฃ Introducing ๐๐๐๐: ๐๐ถ๐ป๐ฒ๐ฎ๐ฟ ๐๐ป-๐๐ผ๐ป๐๐ฒ๐ ๐ ๐๐ฒ๐ฎ๐ฟ๐ป๐ถ๐ป๐ด โจ @CVPR 2026 Oral โจ
๐๐๐๐ trains on videos without manual annotation.
Key idea: An optimal linear mapping that predicts dense cues (e.g. depth, flow), estimated on one video frame, should also predict the corresponding cues of another frame from the same video.
This yields compelling results on dense vision tasks: video object segmentation, (zero-shot) semantic segmentation and surface normal estimation.
Paper, code, models and demo: https://t.co/Xn2SgskKQ8
Joint work with @ma_sundermeyer, Hidenobu Matsuki, David Joseph Tan and @fedassa (and special thanks to David and Federico for hosting my research visit at Google).
#cvpr2026 @Google@MunichCenterML@tumcvg@TU_Muenchen
@matt_barrie@cjoye Even bigger scam is the "world class" universities giving out degrees while their students sleep or work full-time. It's hollowing out the talent pipeline and will cause generations of poverty.
Haven't tried hercules nor have I taken the interview but I feel confident their format is a million times better at selecting engineering ability over leetcoding and making people speak to AI avatars.
I believe we've found the best AI-native coding interview
We call it the โComposer 1 interviewโ
Candidates get 1 hour to build a real, medium-sized project live
The only constraint: they have to use Cursorโs Composer 1 model
@Chip_Prospect@cjoye@FinancialReview A government can either encourage manufacturing and export to other countries or tax the crap out of its own people.
Horne remains true though. The list has companies from US, India, Indonesia and Singapore but few Aussie ones. I don't see much disruption let alone innovation here but will be happy to be proven wrong by new startups.
We officially opened our office in Sydney last week and when we were on the ground, people kept describing Australia as the "lucky country" based on the 1964 book by Donald Horne.
It wasn't a compliment. Horne's point was that Australia got rich on what was in the ground, not what was in people's heads. He postulated that Australia's success came from resource extraction, not technological innovation - and this fear was heightened in the age of AI.
At Sunrise, Rick Baker pointed out that Australia's instinct in this new AI era is to build data centers - the same old playbook:
Build Infrastructure.
Sell the commodity.
Let someone else capture the software margin.
His call to action was a wake up call for ANZ founders to be more ambitious.
Well, from our vantage point at @AnthropicAI, founders from APAC are some of the MOST ambitious:
No matter which part of you Azure look at, one thing is abundantly clear, Azure is a pile of shit in every respect. Terminology, abstractions, documentation, UX, DX, perf or pricing, you name it!
@deedydas I'd rather have a model that collects information and spend its smarts on deducing the steps, than one that's a can of, quickly outdated, recipes.
Large parts of northern and central India are becoming hard to live in.
8 months in a year endure extreme heat, heavy humidity, and pollution that tarnishes even metal and machinery.
Imgine what it's effect must be on human bodies!
People are upset that heaps of money is being made on Polymarket by insider traders betting on tragedies, death and destruction, as if stock markets are any different.