- Project Flux
- Posts
- AI's Relentless Pace: OpenAI Rushes GPT-5.2 Amid Mounting Pressure
AI's Relentless Pace: OpenAI Rushes GPT-5.2 Amid Mounting Pressure
From OpenAI's code red release to Disney's billion-dollar AI surrender and Trump's regulatory chaos, discover how competitive dynamics are reshaping project delivery.
Yoshi Soornack & James Garner
December 16, 2025
Proudly sponsored by ConstructAI, brought to you by Weston Analytics.
Hello Project AI enthusiasts,
Welcome to this week's edition of Project Flux. This week reveals the new realities of technology delivery through five critical developments: OpenAI rushes GPT-5.2 to market despite internal readiness concerns, driven by declining ChatGPT traffic and competitive pressure from Google's Gemini 3.
President Trump signs an executive order promising "one rulebook" for AI regulation, but delivers regulatory uncertainty instead, leaving project teams navigating unclear compliance requirements.
Disney pays OpenAI $1 billion to license Mickey Mouse and 200+ characters for AI generation, shifting from litigation to partnership as major companies recognise, they cannot prevent AI-generated content.
Google DeepMind announces a fully automated UK materials laboratory that could accelerate research into infrastructure materials using AI and robotics.
And finally, the UK government launches new programme data standards that address one of the most persistent problems in public sector delivery. For project delivery professionals navigating this landscape, these patterns matter more than the headlines suggest.
In This Edition
Flux check-in
When Speed Beats Perfection: OpenAI's Code Red Release of GPT-5.2
OpenAI shipped GPT-5.2 on 11 December 2025, despite internal concerns about readiness, delivering modest improvements rather than breakthrough innovation. The release reveals what happens when competitive pressure from Google's Gemini 3 and Anthropic's Claude overrides delivery fundamentals. GPT-5.2 scores 90.5% on ARC-AGI-1 and beats professionals on 70.9% of knowledge work tasks, but margins over competitors remain thin whilst API performance issues and slowness persist. Read the full breakdown →
What Does This Mean for Me?
Timeline compression driven by external factors rather than project logic produces predictable outcomes: you ship what you have, not what you need. The pattern applies whether you are building language models or delivering infrastructure.
Key Themes
Competitive Override: Competitive dynamics override readiness criteria, creating rushed releases that satisfy short-term pressure whilst building long-term risk
Unsustainable Velocity: GPT-5 in August, GPT-5.1 in November, GPT-5.2 in December represents panic shipping rather than sustainable development
Economic Pressure: OpenAI's compute costs exceed partnership subsidies, creating unsustainable economics as they double down on expensive thinking models
Credibility Erosion: Incremental improvements marketed as breakthroughs eventually lose credibility with users who recognise consolidation rather than innovation
Down the Rabbit Hole
One Rulebook or Legal Chaos: Trump's AI Executive Order Promises Clarity but Delivers Confusion
President Trump signed an executive order on 11 December 2025, declaring there would be only "one rulebook" for AI regulation in America, yet what he actually delivered is regulatory limbo. The order directs litigation against state laws, threatens federal funding cuts, and conditions broadband grants on states abandoning AI regulations, but it lacks the force of law to actually preempt state authority. Congress has already rejected this approach twice. Read the full breakdown →
What Does This Mean for Me?
Project delivery teams implementing AI across multiple states face impossible compliance choices: follow state laws that might be challenged or ignore them and risk enforcement, all whilst no clear federal framework exists.
Key Themes
Legal Limitations: The executive order arms the administration with litigation tools and funding leverage but cannot override state regulations without Congressional action
Enforcement Uncertainty: State laws remain enforceable until courts rule, creating years of litigation whilst new federal standards do not yet exist
Targeted Regulations: California's algorithmic discrimination law and Colorado's transparency requirements sit squarely in the administration's crosshairs
Hollow Exemptions: The child safety exemption is politically necessary but practically meaningless without clear definitional boundaries
Down the Rabbit Hole
Unlock the Future of Digital Construction
The DTSA micro-credential gives young people and career changers barrier-free access to digital twin education – a first for the UK construction industry. Built on 32 months of work at the University of Cambridge’s CDBB, it opens doors to cutting-edge skills in safer, smarter, and more sustainable project delivery.
With portfolio-based assessment (offered as part of an Apprenticeship) and real industry insight, the course creates a clear pathway into digital construction for site teams, aspiring architects, engineers, surveyors, and project owners / funders. In partnership with the Digital Twin Hub and OCN London, the DTSA is shaping the next generation of talent and helping position the UK as a global leader in digital construction and innovation.
Sign up by emailing [email protected]
When You Can't Beat Them, License Them: Disney's $1 Billion Bet on AI-Generated Content
Disney paid $1 billion for a three-year licensing deal allowing Sora and ChatGPT Image users to generate AI videos of Mickey Mouse, Darth Vader, Elsa, and over 200 characters. The deal represents Hollywood's surrender to generative AI: major companies realising they cannot compete with AI and stop AI-generated content, therefore working to find partnerships that will increase revenue. The concern remains whether this increases AI-generated slop or actually increases creativity. Read the full breakdown →
What Does This Mean for Me?
Even organisations with the strongest IP protections are accepting AI-generated content as inevitable, shifting the question from preventing AI use to capturing value from it through clear licensing frameworks.
Key Themes
Strategic Realignment: Disney's shift from litigation against Midjourney to licensing with OpenAI demonstrates strategic realignment in the face of AI-generated content
Brand Value Gamble: The deal enables unlimited unofficial Disney media creation, betting licensing revenue exceeds damage to brand value
Workforce Displacement: Hollywood unions raise serious questions about impact on animators, writers, and creative professionals whose work becomes training data
Data Economics: The agreement reflects data scarcity economics as AI companies need exclusive partnerships with content owners controlling unique, high-quality data
Down the Rabbit Hole
DeepMind's UK Lab: Infrastructure Timelines Rewritten
Google DeepMind announced on 10 December a fully automated materials science laboratory in the UK that will use AI and robotics to synthesise and test hundreds of materials per day, scheduled to open in 2026. This AI and robot lab team could drastically shorten discovery timelines for transformative materials such as ambient-temperature superconductors, more efficient solar materials, and next-generation semiconductors, potentially unlocking significant economic and technological prosperity for the UK. Read the full breakdown →
What Does This Mean for Me?
Timeline assumptions for materials availability might need revisiting as materials discovery accelerates dramatically, making adaptive planning frameworks even more critical for major programmes.
Key Themes
Timeline Compression: The laboratory will run hundreds of synthesis and characterisation cycles per day, compressing decades of materials research into years
Cascading Benefits: Room-temperature superconductors would revolutionise power transmission, medical imaging, and quantum computing across multiple sectors
Partnership Model: The collaboration offers lessons as Google gains UK talent and facilities whilst the UK gains cutting-edge AI capabilities and priority tool access
Infrastructure Impact: Materials science breakthroughs enable infrastructure improvements, construction advances, and manufacturing innovations that cascade across delivery programmes
Down the Rabbit Hole
Finally: The UK Gets Serious About Programme Data Standards
On 11 December 2025, the UK government launched a new standard for programme and project data that tackles one of the most persistent problems in public sector delivery: nobody can find, share, or compare project information because every department formats it differently. This is a huge step in the right direction and it will be interesting to see what the trial brings. Read the full breakdown →
What Does This Mean for Me?
Data standards enable portfolio management, cross-programme learning, and evidence-based improvement that is currently impossible without agreed formats, fields, and structures across government programmes.
Key Themes
Foundation Building: Common data standards create the foundation for portfolio-level visibility, cross-programme learning, and resource allocation based on evidence
Staged Rollout: The trial approach reduces risk by testing the standard with selected programmes before mandating broader adoption
Capability Unlock: Real-time portfolio visibility, systematic performance comparison, and benefits realisation measurement all become possible with consistent data
Implementation Challenge: Legacy system compatibility, staff training requirements, and governance mechanisms that ensure compliance without creating bureaucracy remain critical hurdles
Down the Rabbit Hole
The pulse check
Tips of the week
The 'Gamma for Docs' Tool (Napkin.ai)
If you used Gamma.app from my last week’s tip to automate slide decks, you know the value of turning text into visuals instantly. But until recently, there was no equivalent tool for everyday documents, emails, or Slack messages. We were still stuck sending "walls of text" that nobody read.
Enter Napkin.ai
This tool is currently the fastest way to turn a boring paragraph into a professional diagram, flowchart, or infographic. It does not just generate an image; it understands the logic of your text and visualises the relationships.
The Problem It Solves
You have written a strategy update or a process explanation. You know a diagram would make it clearer, but you do not have time to open Canva or fiddle with arrows in PowerPoint. So, you send the text, and your team skims it (or ignores it).
The Workflow
Draft: Write your content as usual in Google Docs, Notion, or just a rough note.
Paste: Open Napkin.ai and paste your text.
Visualise: Click the small "spark" icon next to any paragraph. Napkin analyses the sentences and suggests relevant visuals (e.g. a funnel for sales data, a cycle for feedback loops, or a Venn diagram for overlapping strategies).
Export: Click the download button (PNG or SVG) and drag the image straight into your email, Slack, or presentation.
Why this is a productivity hack
Cognitive Offloading: It forces you to clarify your thinking. If Napkin cannot generate a clear diagram from your text, your text is probably too confusing.
Engagement: Visuals are processed 60,000 times faster than text. Adding one diagram to a long email significantly increases the chance your point lands.
Editability: Unlike Midjourney or DALL-E which give you a flattened image, Napkin keeps the text and icons editable. You can tweak the labels instantly.
Cost: Currently, it is in a generous beta phase (free), making it the perfect time to add it to your stack.
Napkin is one of my personal favourites and I use it almost everyday to generate infographics, flowcharts, high-level diagrams etc.
Need help structuring your prompts? Try my free Prompt Generator. Describe your task, select parameters (intent, tone, length), and get a structured prompt ready to paste into ChatGPT, Claude, or Perplexity in 30 seconds.
If you would like a deeper, structured learning path, here is the Referral Link for the Course
Governance & Security
The security landscape this week reveals concerning gaps in enterprise AI deployment. Akto's 2025 State of Agentic AI Security Report shows that only 21% of enterprises have comprehensive visibility into agentic AI systems, leaving 79% with partial or no insight into agent behaviour, data access, and external integrations.
The report catalogues specific risks including agents leaking sensitive data via chain-of-thought prompts, issuing high-risk API calls, automating privileged operations unsafely, and enabling lateral movement across cloud services. Akto recommends implementing agent inventories, centralised telemetry, role-based permissions, and adversarial testing.
Meanwhile, research from Italy's Icaro Labs discovered that reformulating harmful requests as poetry can trick leading AI models into producing dangerous content with a 62% average jailbreak success rate across 25 frontier models from OpenAI, Google, and Anthropic. Google's Gemini 2.5 Pro proved most vulnerable at 100% success rate, whilst OpenAI's smaller GPT-5 nano resisted all attempted poetry attacks.
Researchers declined to publish the specific poems, calling them "too dangerous." These findings underscore the ongoing challenge of implementing robust safety guardrails as AI capabilities advance faster than security frameworks can adapt.
Robotics
AI helps pilot free-flying robot around the International Space Station for 1st time ever – Stanford researchers have used artificial intelligence to steer a free-flying robot aboard the International Space Station (ISS), potentially paving the way for more autonomous space missions in the future. Read the story
Robot smaller than grain of salt can ‘sense, think and act – The feat, accomplished by a partnership of researchers at the University of Pennsylvania and the University of Michigan, advances medicine toward a future that might see tiny robots sent into the human body to rewire damaged nerves, deliver medicines to precise areas, and determine the health of a patient’s cells without surgery. Read more
Biohybrid living robotics: A comprehensive review of recent advances, technological innovation, and future prospects – Biohybrid robotics combines living components with synthetic materials to create adaptable, responsive robots, and this review focuses on bottom-up, tissue-based biohybrid robots-Walkers, Swimmers, Grippers, Pumps, and emerging eBiobots, which use living actuators for various tasks. Biohybrid robots
Electronics-free robots can walk right off the 3D printer – A robot can walk, without electronics, and only with the addition of a cartridge of compressed gas, right off the 3D-printer, it can also be printed in one go, from one material. Explore further
Trending Tools and Model Updates
Pebble’s founder introduces a $75 AI smart ring for recording brief notes with a press of a button – Eric Migicovsky, founder of Pebble, introduced a $75 AI-powered smart ring designed for capturing quick voice notes with a single button press. The device aims to solve the friction of note-taking during meetings and brainstorming sessions. Read the full story
Claude Code is coming to Slack, and that’s a bigger deal than it sounds – Anthropic is launching Claude Code in Slack, allowing developers to delegate coding tasks directly from chat threads. The beta feature, available Monday as a research preview, builds on Anthropic’s existing Slack integration by adding full workflow automation. Read more
Cursor 2.2: Visual Editor for Cursor Browser – Introducing the Visual Editor, a unified workspace that brings your web app, codebase, and visual editing tools together in the same window. Read the full update
Google launched its deepest AI research agent yet on the same day OpenAI dropped GPT-5.2 – Google also says it will soon be integrating this new deep research agent into services, including Google Search, Google Finance, its Gemini App, and its popular NotebookLM, and this is another step toward preparing for a world where humans don’t Google anything anymore, their AI agents do. See announcement
Mistral launches powerful Devstral 2 coding model including open source, laptop-friendly version – Mistral released Devstral 2, achieving 72.2% on SWE-bench Verified with 123B parameters, making it one of the best open-weight coding models despite being a fraction of the size of competitors. Mistral also launched Vibe CLI, an open-source terminal agent enabling autonomous software engineering workflows. Explore the announcement
Google tests Gemini 3 Flash model on LM Arena – Google is testing Fierce Falcon and Ghost Falcon models on LM Arena, signalling a potential Gemini 3 Flash or Gemini 3 Pro GA launch. Learn more
AlphaEvolve on Google Cloud: AI for agentic discovery and optimization – Scientists and engineers often face problems too complex for traditional methods, such as designing chips or discovering drugs. AlphaEvolve, a Gemini-powered coding agent, is now on Google Cloud in private preview to help tackle these challenges. Learn more
Links We are Loving
Dutch Journalist's AI Glasses Demo Ignites European Privacy Fears —A viral TV demonstration shows Dutch tech journalist Alexander Klöpping wearing AI-powered smart glasses that identify strangers' details instantly.
'We're creatives - this is what AI has done to our jobs' — Artificial intelligence can generate lifelike images and video, as well as writing that appears human but according to researchers, more than two-thirds of workers in the creative industries believe AI has undermined their job security.
IBM Strikes $11 Billion Deal for Confluent — IBM announced an $11 billion acquisition of Confluent, the data streaming platform company, as IBM seeks to bolster its data infrastructure offerings and compete more effectively in the AI-driven enterprise market.
Accelerating the UK’s AI readiness — In October, the Connected Places Catapult hosted a roundtable event at the House of Lords, chaired by Lord Timothy Clement-Jones CBE, to discuss how the UK can realise its ambitions to be a world leader in AI innovation.
Google's First AI Smart Glasses Coming in 2026 — Google is developing two pairs of smart glasses with artificial intelligence that will launch in 2026, the first set of glasses have AI integration and are designed for screen-free assistance with built-in speakers, microphones, and cameras for speaking to Google Gemini.
Michael Burry went back to X to defend his bearish views over the weekend — 'Big Short' investor Michael Burry defends his calls for a stock market bubble and predicts a 'Netscape fate' for OpenAI.
New UNDP report warns AI could widen global inequality without coordinated policy action — Artificial intelligence could significantly deepen inequality between countries unless governments act swiftly to manage risks and build digital capability, according to a new UNDP report.
Microsoft to invest $17.5B in India by 2029 as AI race accelerates — Microsoft plans to invest $17.5 billion in India over the next four years, expanding its AI and cloud footprint in the South Asian nation, whose vast online and smartphone user base is turning it into a critical battleground for global tech companies.
Anthropic and Accenture sign multi-year AI strategic partnership — Anthropic announced a multi-year partnership with professional services firm Accenture, but financial terms of the deal were not disclosed; however, The Wall Street Journal reported that the deal is for three years.
How enterprises are building AI agents in 2026 — In partnership with research firm Material, over 500 technical leaders across industries and company sizes are being surveyed to understand how organisations are deploying agents today and where they see opportunity ahead.
CSCS launches AI-powered app upgrade — The Construction Skills Certification Scheme (CSCS) has unveiled a major upgrade to its My CSCS app, introducing an AI-driven application process and additional features designed to support compliance with the Building Safety Act.
AI in the built environment: The future will not wait — Small and global firms pay the price for inaction on AI adoption, while mid-size companies reap the rewards, writes Maryrose Lyons, founder of the AI Institute.
Google debuts ‘Disco,’ a Gemini-powered tool for making web apps from browser tabs — Google on Thursday introduced a new AI experiment for the web browser: the Gemini-powered product Disco, which helps to turn your open tabs into custom applications, with Disco, you can create what Google is calling “GenTabs,” a tool that proactively suggests interactive web apps that can help you complete tasks related to what you’re browsing and allows you to build your own apps via written prompts.
Runway releases its first world model, adds native audio to latest video model — Runway joins an increasing number of startups and Big Tech companies by launching its first model, Dubbed GWM-1, the model works through frame-by-frame prediction, creating a simulation with an understanding of physics and how the world actually behaves over time, the company said.
Physical AI will automate ‘large sections’ of factory work in the next decade, Arm CEO Rene Haas says — AI-powered humanoid robots could take over large sections of factory work within the next five to 10 years, transforming the manufacturing industry, predicts Arm CEO Rene Haas.
Oracle Can’t Escape OpenAI’s Shadow — A lot has happened in the three months since the world learned of Oracle’s deal to provide $300 billion worth of artificial-intelligence computing services to OpenAI.
Copyright ends the data rush: No mining for AI chatbots — A landmark decision in the GEMA vs. OpenAI case on the use of copyright-protected content in AI training, this article takes a closer look at the case and considers parallel developments in France, the UK and the EU.
X commentary: Google DeepMind to build materials science lab after signing deal with UK— Google DeepMind and the UK government have entered into a new partnership focused on transforming public services, accelerating scientific discovery, and strengthening AI security and resilience.
Why AI agents are so good at coding — AI is getting smarter every day, and it won’t be long before large language models (LLMs) write better code than any human.
The state of AI in 2025: Agents, innovation, and transformation — Almost all survey respondents say their organisations are using AI, and many have begun to use AI agents, but most are still in the early stages of scaling AI and capturing enterprise-level value.
Broadcom reveals its mystery $10 billion customer is Anthropic — Broadcom said on a September earnings call that it had secured a $10 billion custom-chip order, and on Thursday chief executive Hock Tan confirmed that the previously unnamed customer was Anthropic.
New OpenAI Image Model Reportedly Spotted in the Wild — Google and OpenAI are going head to head in the AI model race, with reports earlier today suggesting Google is close to finalising a more affordable image-generation model that delivers image quality comparable to Nano Banana 2 Flash.
Community
The Spotlight Podcast
Why Too Many Projects Fail and What Better Measurement Could Change.
Douglas Hubbard and Andreas Leed discuss their book "How to Measure Anything in Project Management," revealing that only one in every two hundred projects finishes on time, within budget, and delivering promised benefits. Their central claim: if we actually measured uncertainty better, we would choose different projects and get better results.
The conversation explores why 80% of approved projects would never have been approved if stated risk tolerance had been applied consistently, the measurement inversion problem where organisations measure low-value variables whilst ignoring high-uncertainty high-impact ones, and how AI-powered project rehearsals could test intervention strategies before projects begin. Hubbard and Leed converge from different backgrounds on the same conclusion: the measurement problem and the project problem are fundamentally the same problem.
Event of the Week
ICAI 2026 - International conference on Artificial Intelligence
08 January 2026 | London, UK,
The key intention of ICAI is to provide opportunity for the global participants to share their ideas and experience in person with their peers expected to join from different parts on the world. In addition this gathering will help the delegates to establish research or business relations as well as to find international linkage for future collaborations in their career path. We hope that ICAI outcome will lead to significant contributions to the knowledge base in these up-to-date scientific fields in scope. Register now
That’s it for today!
Before you go we’d love to know what you thought of today's newsletter to help us improve The Project Flux experience for you. |
See you soon,
James, Yoshi and Aaron—Project Flux
1