hello!
Cognitive Warmup. The man many credit with leading the AI revolution (I certainly don’t; refrain from getting started on that), Sam Altman, recently said that if AI eliminates some jobs, they probably weren’t even “real jobs” to begin with. What is Altman talking about? Law enforcement? medical professionals? Lawyers and legal experts? Journalist? Creative professional? the mailman? I’m sure CEOs and their chosen favorites will be safe from the massive AI experiments that enterprises around the world are slowly burning their fingers with. This is the same Sam Altman who was recently asked on a podcast how his $13 billion nonprofit claim could support $1.4 trillion in spending commitments, and he went on the defensive, blaming some non-existent short sellers. Keep in mind, there has been speculation about an OpenAI IPO for some time. What if tomorrow an AI decided Sam Altman is redundant? Altman-In reality it should be peaceful.
algorithm
This week, we talk about another research paper that exposes AI’s security nightmare, explains what a ‘free’ AI subscription plan really means, and what Figma’s latest acquisition means.
LLM security, still unresolved
A new paper titled The attacker is second: Strong adaptive attacks bypass defenses against LLM jailbreaks and quick injections ,You can read it here), confirms an inconvenient truth – the issue of large language model security is far from resolved. Researchers from Northeastern University, ETH Zurich, as well as AI companies Google DeepMind, OpenAI, and Anthropic, tested 12 robust jailbreak and prompt-injection defenses, but found that they all collapsed when faced with both human- and LLM-powered adaptive adversaries.
This study emphasizes that static evaluation, where models are tested against a fixed set of attacks, does not reflect real-world dynamics. Once attackers adapt to changing guardrails, the success rate increases to more than 90%. In short, security layers remain brittle. Seismic, if you refer to it as such, with the widespread personal and business use of LLMs under the guise of AI chatbots.
The real cost of “free” Gemini AI for Jio users
There was a time when data was the new oil. Now the blind race is to make AI everything…oil, data, gold, whatever you want to make from it. Indian telecom companies are rushing to bundle AI in the hope that it will increase average revenue per user (ARPU). This is the holy grail for any telecom player since I have been following the sector since the late 2000s. Following the excitement caused by the Airtel and Perplexity bundles a few weeks ago, I would note that the social media excitement (usually led by thread enthusiasts) around the Reliance Jio and Google Gemini AI Pro bundles was getting a little higher. As generous as it sounds on paper (the 18-month bundle beats the 12-month bundle of Airtel’s Perplexity plan), you have to factor in a cost at the end of the period.
At this time, Jio says the early access is available to users on “unlimited 5G plans”, and is aimed mostly at the young demographic of students. The most affordable 5G plan works among these unlimited 5G plans. Rs.349 with 28 days validity and 2GB data per day. This simply means that over an 18 month period, you are spending 6,282 (ie 349×18; you will be better off 3,999+ 1,799 plans instead, if there is certainty about the longer term) or remain eligible for this offer. There’s the cost, the companionship aspect (Jio will surely find your ‘loyalty’ invaluable) and the need to recharge regularly (one skip, and the offer breaks). Smart marketing and smart math, but you have to do the work if the calculations work for you. After that 18 month period, all the illusion of ‘free’ will also disappear, and you will be left with almost no money spent. 19,000 per year because Google Drive storage will inevitably fill up way beyond the 15GB free limit and you will have handed over all your thinking to AI.
Figma’s VV step towards a creative engine
Figma has acquired VV, an AI-powered node-based (this means a visual, flowchart-like interface where creative processes and effects are listed individually) media generation platform that blends image, video, and animation tools. This signals Figma’s intention to evolve beyond static design into full-spectrum generative creation. VV’s technology enables multi-model composition, giving designers the ability to mix generative models, tweak parameters, and create layered, editable results instead of one-shot outputs. Think of it as a creative workspace that is evolving into something where design, iteration, and storytelling co-exist. The future of design is probably not about replacing imagination, but about making it more editable.
Last time on Neural Dispatch: The Lack of Intelligence in Agentic AI, a DeepSeek Moment, and Nvidia’s AI Supercomputer
ready
Adobe Premiere comes to the iPhone, and editing on a mobile device finally becomes truly powerful. Much more than those Instagram reels focused apps that litter the App Store.
Adobe has quietly accomplished a lot here. A proper version of Premiere on iPhone. This is not the usual cool, casual, and quirky mobile-first editing app that many of the Reels crowd-pleasers seem to attract. Adobe is no stranger to turning out powerful mobile apps that don’t compromise capabilities — Lightroom is a personal favorite, as is Project IndigoSomething I wrote about recently. It lives up to the Premiere Pro name, and so includes multi-track editing, layered audio, full-resolution 4K HDR export, AI-powered background removal, and animated captions that can actually sync with your voice. In short, this is not a toy for reels, but a full-fledged editing suite. It is expected that it will get even more powerful features in the future. For now, it’s free to download and edit, with no limitations or features pushed behind a subscription wall. But if you want to use generative AI features you will have to pay for Firefly for credits.
This step makes perfect sense. If your workflow already lives on your phone (i.e. shooting, scripting, and posting) dragging footage to the desktop every time will feel like a creative tax. The ability to color-grade, trim, add text overlays, and even clean up your audio (Adobe Podcast talent has made its way here), all of which add credibility to Adobe Premiere’s push for relevancy. The app also syncs seamlessly with your desktop Premiere, just in case you want to pick up where your thumb left off. Android phone users, if you’re wondering where all this leaves you, Adobe says an Android version of Premiere is also planned at some point. keep waiting.
Be sure to check out my other newsletter, Wired Wisdom: Adobe’s AI Avenue, Asus’s arrogant console pricing, and thoughts on brain rot
Thinking
“Creativity is the balance of innovation and imagination, technology and humanity. We will provide tools, platforms and integrate with the ecosystem to empower creators, unleashing their most daring imagination. We are blending AI in simple and intuitive ways.” – Shantanu Narayan, CEO of Adobe, in Max 2025.
Its a statement This goes beyond Adobe’s product playbook. This is a worldview, a balanced view, at a time when AI companies are talking big numbers and making even bigger promises. Narayan’s emphasis on ecosystems, not individual devices, defines where at least creative-focused technology is headed. The future of creativity may lie in how seamlessly human intentions blend with AI, not how powerful AI is at making decisions on behalf of humans. For anyone crafting in this era – be it journalists, designers, filmmakers, coders – the real question is not “Can AI do this?” But “Can my ecosystem express me faster, freer and better?” As Narayan says, this is where technology meets humanity.
Context: At Adobe MAX 2025, Shantanu Narayan reframes the creative industry’s AI conversation. This is not a fight algorithm supremacy But to find a good balance between imagination and technology. Their ecosystem position finds an example within Adobe’s own suite, from Firefly to Premiere to Express, which should be able to behave as a adaptive, creative A platform that allows users to talk to each other when needed. In Narayan’s framing, the company’s value is not in AI itself, but as a connective tissue that lets humans and machines co-create with less friction.
A reality check: While it’s easy to romanticize that balance of AI and simplicity, the real challenge is maintaining authorship and creative authenticity as automation deepens. The more generative systems assist, the more they risk blurring ownership and originality – especially when AI handles the production as well as the polish. Adobe’s approach will depend on whether creators feel empowered By their agentic devices, or feel sidelined. True creativity thrives not only on accessible technology, but also on maintaining the distinctive human fingerprint, insists Narayan. Time will tell us whether Adobe’s approach succeeds, or whether changes will need to be made.
Neural Dispatch is your weekly guide to the rapidly evolving landscape of artificial intelligence. Each edition provides curated insights on the critical technologies, practical applications and strategic implications shaping our digital future.
