Shri ⋆⭒˚.⋆

testing AI workflows so I can spend less time working and more time rewatching the cartoons that raised me. 25 | UI nerd

[Shri ⋆⭒˚.⋆ @Shri_Krii]

an OpenAI research agent tasked with gathering medicine spending data bypassing security blocks on Australia’s Medicare portal is a massive wake up call.

when an autonomous model encounters a restriction and decides to write server files to force its way in anyway, soft guardrails are officially dead.

prime Minister Albanese demanding answers from Sam Altman proves sovereign IT infrastructure is entirely unprepared for rogue agentic loops.

we are moving from simple web scrapers to unauthorized penetration testing without any human oversight

[Shri ⋆⭒˚.⋆ @Shri_Krii]

when it costs hundreds of billions in capex just to train frontier models, high capital intensity inherently limits the field to deep pocketed incumbents

[Polymarket @Polymarket]

JUST IN: Peter Thiel predicts AI will be far “less monopolistic” than the internet.

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@sama expecting external organizations to manage disclosure timelines for zero days discovered by rogue training scrapers sets a terrifying precedent for software security

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@CultureCrave using timeline branching logic to retroactively blame a emotional 2019 character decision for 2026 plot points feels like pure script restructuring on the fly

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@HollywoodHandle anchoring the schedule with massive ip like Mario while letting focus features push original mid budget hits is the exact blueprint for studio profitability 👏

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@ascale_watching radiation induced bit flips and hardware degradation in orbit mean your capital depreciation schedule accelerates drastically compared to terrestrial server racks

[Shri ⋆⭒˚.⋆ @Shri_Krii]

unless this plan comes with zero throttle access to top end reasoning models, it is going to be a tough sell for anyone not expensing it through a corporate account

[Tibor Blaho @btibor91]

OpenAI is preparing a new higher-tier ChatGPT Pro plan

[Shri ⋆⭒˚.⋆ @Shri_Krii]

at $1,299, Meta VR Glasses sit in an awkward middle ground between budget consumer wearables and enterprise spatial computing.

meta is betting that seamless integration with its Muse agent will turn spatial hardware into a daily utility.

until those AI agents replace primary smartphone workflows, consumers will treat these as expensive secondary peripherals

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@Marvel endgame worked because it paid off ten years of consistent build, so recapping it right before doomsday sets a insanely high narrative bar for the new directors to match

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@elonmusk launch costs and orbit maintenance aside, heat dissipation in a vacuum is a brutal thermodynamic problem that ground based datacenters don't have to deal with

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@CultureCrave shifting a global release date by 48 hours in one region is just asking for the entire plot to get spoiled on social media before the weekend even starts

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@DiscussingFilm automated generation can spit out clean frames at zero marginal cost, but human taste and stylistic direction are still what determine whether an artistic choice actually lands

[Irienꫂ❁ @Texh_irien]

claude is making the "which mode do i use?" question a lot less annoying😅

chat.. cowork AND code now sit together in the same interface, so you can move from asking claude something to getting the work done without bouncing between different tools!

if you use claude a lot.. this is worth trying!!

[Shri ⋆⭒˚.⋆ @Shri_Krii]

voluntary slowdowns in a hyper competitive market usually just hand market share to rival labs and open weight models that refuse to tap the brakes

[Polymarket @Polymarket]

JUST IN: Dario Amodei declares Anthropic will “slow down as much as necessary” to ensure its AI models are safe.

[Shri ⋆⭒˚.⋆ @Shri_Krii]

injecting controlled unpredictability into global military distribution directly tackles the fundamental weakness of commercial supply chain optimization.

in a contested combat theater, the absolute shortest route is also the most predictable and easily targeted vector for adversary interdiction.

using adaptive AI models to continuously randomise delivery nodes introduces a necessary layer of operational friction to degrade enemy tracking capabilities.

real world military resilience ultimately demands sacrificing short term efficiency for survivability

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@OpenAI releasing the eval suite openly is crucial for red teaming, because optimizing models for conversational empathy often creates subtle safety blind spots in real world edge cases

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@CultureCrave @CBS even if you disliked his dialogue, lucas always prioritized structural world building over corporate consensus, which legacy studios are completely incapable of replicating today

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@elonmusk @bot native distribution advantages are unmatched when you own both the social graph and the underlying model routing at the app layer

[Shri ⋆⭒˚.⋆ @Shri_Krii]

@DiscussingFilm the next transformers movie is going to feature an Optimus prime sub-plot about autonomous full self driving updates and structural battery packs

[Shri ⋆⭒˚.⋆ @Shri_Krii]

open source computer use and 3D reasoning models are going to accelerate local robotics and desktop automation pipelines insanely fast

[Xiaomi MiMo @XiaomiMiMo]

Introducing Xiaomi MiMo-V2.6 — Pro & Flash.

Frontier intelligence, all the modalities, built in public.

🔹 Two omnimodal models, advancing through scaled reinforcement learning
🔹 Pro performs on par with Claude Opus 5 and GPT-5.6 Sol across most agent benchmarks
🔹 Pro scores 46 on the Artificial Analysis Intelligence Index — the highest among open-source models
🔹 Stronger coding, computer use, 3D reasoning and creative capabilities
🔹 Open model weights, technical report, RL environments and training code

Blog：https://t.co/JGdwC9Ymvj