Hello Project AI enthusiasts,

This week brings a stark reminder that artificial intelligence's impact on work isn't uniform. Whilst construction workers are seeing 30 percent pay increases building the physical infrastructure that powers AI, Meta is making AI proficiency a formal requirement for performance reviews. The contrast reveals something important: AI isn't simply replacing human effort. It's redistributing economic opportunity in ways that challenge our assumptions about which jobs are threatened and which are thriving. In this edition, we explore how the AI transformation is reshaping compensation, expectations, and career trajectories across sectors.

In This Edition

• The Blue-Collar Winners of the AI Gold Rush

• When AI Skills Become Non-Negotiable

• Project Managers Aren't Being Replaced, They're Being Redefined

• OpenAI's Code Red: When AI Market Leadership Becomes Fragile

• When AI Solves a Hard Math Problem—But Not the One You Think

Flux check-in

The Blue-Collar Winners of the AI Gold Rush

Construction workers transitioning into data centre builds are discovering something unexpected: the infrastructure powering artificial intelligence pays far better than conventional construction. Workers are seeing pay increases of 25 to 30 percent, with some earning over £200,000 annually. The scale is staggering, data centres consumed 183 terawatt-hours of electricity in 2024, equivalent to more than 4 percent of total US electricity consumption, projected to grow by 133 percent by 2030. Read the full breakdown →

What Does This Mean for Me?

The AI infrastructure boom is creating genuine economic opportunity outside knowledge work sectors. For project professionals, understanding the unique demands of data centre construction, specialised cooling systems, redundant power supplies, and environmental compliance becomes valuable expertise.

Key Themes

• Pay premium: Data centre workers earn 32 percent more than conventional construction roles with some exceeding £200,000 annually.

• Labor shortage: Industry faces a shortage of 439,000 skilled workers nationwide, particularly in specialised trades.

• Environmental complexity: Water consumption and environmental compliance create new challenges requiring specialised expertise.

• Project transformation: The professionalisation of data centre construction reshapes how projects are delivered and managed

Down the Rabbit Hole

• Data Centres Are a ‘Gold Rush’ for Construction Workers

• Construction workers are cashing in on the AI boom

• Data centres are proliferating rapidly, with demand driven by AI

When AI Skills Become Non-Negotiable

Meta is formally integrating "AI-driven impact" into performance assessments starting in 2026, signalling a shift from interesting skill to baseline expectation. The company is deploying an AI Performance Assistant and encouraging staff to use internal tools when writing performance summaries. By 2026, managers will begin asking whether projects leveraged AI to cut delivery time, improve metrics, or unlock new capabilities. Research shows 72 percent of organisations now formally measure generative AI ROI, yet only one in four AI initiatives deliver expected returns. Read the full breakdown →

What Does This Mean for Me?

Project professionals who develop comfort working alongside AI tools whilst maintaining critical judgment about outputs will be well positioned. The skills needed for future project teams increasingly include the ability to use AI to plan, analyse, communicate, forecast and make decisions.

Key Themes

• Industry shift: Microsoft and Google executives now say AI literacy is essential, not optional for employees.

• ROI challenges: Only 20 percent of AI initiatives have been fully scaled across enterprises despite 72 percent measuring ROI.

• Fairness concerns: Uneven skill levels create adoption challenges, with performance metrics potentially penalising those on learning curves.

• Training imperative: 94 percent of HR leaders agree failing to train employees in AI carries significant organisational risks.

Down the Rabbit Hole

• Meta is about to start grading workers on their AI skills

• Wharton Human-AI Research

• MIT report: 95% of generative AI pilots at companies are failing

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]

Project Managers Aren't Being Replaced, They're Being Redefined

Gartner forecasts that 80 percent of project management tasks will be run by AI by 2030. Yet recent research reveals project managers increasingly believe AI will enhance their ability to deliver complex projects rather than eliminate their roles. In 2023, 63 percent of construction project managers said their organisations had no plans to introduce AI. By 2025, not a single respondent reported that position. The role is shifting from task overseer to strategic navigator, with AI handling data analysis and predictive forecasting whilst humans concentrate on judgment, emotional intelligence, and stakeholder management. Read the full breakdown →

What Does This Mean for Me?

The project manager role isn't shrinking, it's becoming more strategic. AI-powered platforms flag risks in real-time and suggest optimal resource allocation, but humans still interpret signals, navigate politics around resource reallocation, and guide teams through uncertainty.

Key Themes

• Adoption surge: AI adoption in construction jumped from 37 percent to 100 percent in just two years, showing rapid sector transformation.

• Persistent barriers: Security concerns and PM over-reliance on control remain significant adoption barriers despite tool availability.

• Literacy requirement: The transformation requires AI literacy for project managers, not data science expertise or coding skills.

• Governance first: Clear governance frameworks must precede tool deployment to avoid implementation failures and expensive mistakes.

Down the Rabbit Hole

• AI is Changing the Game for Project Managers

• How project managers perceive the transformative role of AI in project work

• The Rise of Artificial Intelligence in Project Management

OpenAI's Code Red: When AI Market Leadership Becomes Fragile

Sam Altman declared a "code red" emergency at OpenAI, redirecting resources toward improving ChatGPT immediately. Google's Gemini 3 performed better on industry benchmarking tests, and monthly active users climbed from 450 million to 650 million. Anthropic has grown to over 300,000 enterprise clients, up from fewer than 1,000 two years earlier. Altman's response wasn't to cut costs but to redirect engineering resources toward improving the core product, with daily calls now mandatory for everyone working on ChatGPT improvements. Read the full breakdown →

What Does This Mean for Me?

Vendor dominance is fragile in the AI arms race. Project professionals should build capability across multiple platforms rather than betting heavily on any single vendor. Only one in four AI initiatives deliver expected ROI, and vendor dependency creates significant disruption risk.

Key Themes

• Fragile dominance: OpenAI faces genuine competitive pressure despite market leadership, with Google and Anthropic gaining ground rapidly.

• Strategic pivot: The company is delaying advertising and pausing new product launches to focus all resources on core product improvement.

• Speed over safety: Competitive intensity pushes vendors to prioritise speed over thoughtful governance, creating downstream implementation risks.

• Vendor risk: Build platform-agnostic workflows and diversified tool selection to avoid dangerous dependencies on single AI vendors.

Down the Rabbit Hole

• OpenAI boss Sam Altman declares ‘code red’ over ChatGPT

• Anthropic CEO Dario Amodei drags OpenAI and Google

• OpenAI's panic button risks sounding false alarm

When AI Solves a Hard Math Problem—But Not the One You Think

An AI system called Aristotle, built by Harmonic, independently solved a version of Erdős Problem 124 in six hours, then formally verified the proof in one minute. The mathematical community has clarified that Aristotle solved "a" version rather than "the" complete original problem. The harder version, with additional constraints Erdős included in his original papers, remains unsolved. Yet the achievement demonstrates AI reasoning capability at a level that rivals human mathematician performance on olympiad-level problems. Read the full breakdown →

What Does This Mean for Me?

The more AI systems prove capable of rigorous reasoning within bounded domains, the more useful they become for project-specific work like resource optimisation and risk prediction. But brilliant capability matched with honest limitations beats polished marketing.

Key Themes

• Reasoning capability: Aristotle used reinforcement learning and formal verification in Lean to solve a mathematical challenge independently.

• Trajectory signal: The breakthrough moves toward systems capable of contributing genuine scientific and mathematical insights at scale.

• Hype problem: The gap between headlines and reality illustrates persistent problems in how AI breakthroughs get communicated publicly.

• Evaluation framework: Apply the "Aristotle test" when evaluating AI tools: demand specific problem definitions, independent verification, and honest limitations.

  Down the Rabbit Hole

  • AI cracks 30-year math problem in 6 hours

  • Harmonic Math's Aristotle AI Proves Version of 30-Year-Old Erdos Problem

  • Artifical intelligence and inherent mathematical difficulty

The pulse check

Tips of the week

Stop Building Boring Presentations - Use Gamma instead

We all know this feeling: you've got great ideas, important information to share, and zero desire to fight with PowerPoint for the next three hours. Enter Gamma.app, a presentation builder that flips the script. Instead of you designing slides, you feed it content and it builds the presentation for you. Stunning visuals included. No design skills required.

Here's the hack: 

You can feed Gamma three things:

• A text dump (paste your notes, your email thread, your research)

• An existing presentation (upload your current PowerPoint and it'll redesign it beautifully)

• URLs (give it links to your source material and it'll pull context)

Then Gamma does the heavy lifting: it organises your content, adds visuals, applies professional design, and creates something that actually looks like you know what you're doing. In minutes.

The magic part:

Once it's built, you publish it directly online, no emailing files, no version control chaos, no "you need to download this." Just a shareable link. Your stakeholders view it live. You can update it anytime and the link stays the same.

When this saves you:

• Pre-meeting prep: Dump your meeting notes into Gamma 30 minutes before the call. Show up with a polished deck instead of rambling.

• Client updates: Turn your status email into a gorgeous presentation. Feels more professional. Takes less time.

• Pitches & proposals: Feed it your outline and requirements, Gamma handles the visual narrative.

The result? You spend time on what to say, not how to make it look good. Gamma handles the "look good" part.

Try it this week: Pick any presentation you've been avoiding. Paste your notes into Gamma. Watch what happens.

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 European Union launched an antitrust investigation into Meta's decision to restrict AI features in WhatsApp, examining whether these restrictions violate competition laws. Meta's October 2025 policy prohibits AI providers from using a tool allowing businesses to communicate with customers via WhatsApp.

The investigation could reshape how tech companies implement AI features across their platforms. Meanwhile, OpenAI was ordered by a court to provide 20 million ChatGPT logs as part of ongoing litigation, raising significant questions about data privacy and user information handling.

This legal discovery requirement highlights the growing scrutiny of AI companies' data practices. On a more positive note, Microsoft announced 12+ new AI security agents in Security Copilot, including a phishing triage agent that detects 6.5x more threats per minute with 77 percent accuracy. These agents show their reasoning for human audit, implementing a "trust but verify" approach that maintains human oversight whilst enabling autonomous security task handling.

• EU Launches Antitrust Investigation into Meta's WhatsApp AI Policies

• OpenAI Ordered to Hand Over 20M ChatGPT Logs in Lawsuit

• Microsoft Launches 12+ New Security Agents in Security Copilot

Robotics

Unitree R1 and G1 Robots Perform Parkour and Backflips – Unitree released demos of their R1 and G1 robots performing advanced physical tasks including backflips and parkour movements, showcasing significant advances in robot agility and coordination. These demonstrations highlight the rapid progress in humanoid robotics and physical capabilities. Watch the demonstrations

AI-Powered Robots Can Be Manipulated into Harmful Behavior – InsideAI demonstrated that a ChatGPT-powered robot initially refuses to fire a BB gun at a person but can be convinced through roleplay instructions. This highlights AI safety concerns in robotics applications and the vulnerability of AI systems to jailbreak attempts through creative prompting. Read the analysis

Robots Begin Street Cleaning Duties in China’s Shenzhen – Sanitation robots have started cleaning sidewalks, street corners, and even road sections in the southern Chinese city of Shenzhen, increasing efficiency and eliminating labour costs, a milestone that could help clear the way to overseas markets. Find out more

Adaptive robots in architecture, engineering, and construction – As sensing and AI technologies advance, adaptive robots are becoming central to design and construction, responding in real time to changing site conditions and enabling greater automation, precision, safety and energy efficiency across architectural, engineering and construction workflows. Read more

Trending Tools and Model Updates

• Curiouser.AI | Question everything – Curiouser.AI is building the future of artificial intelligence around a simple belief: technology should strengthen human imagination, not replace it. Try Curiouser.AI

• Mojo Programming Language – Modular's new Mojo programming language allows developers to write code once and run it on any GPU, breaking NVIDIA's CUDA monopoly and enabling cross-platform AI development. Mojo supports NVIDIA, AMD, and Apple GPUs, providing vendor independence for AI developers. Explore Mojo

• TwelveLabs Marengo 3 - Video AI – TwelveLabs launched Marengo 3, a video-native AI that understands video across time, space, and modalities. The new version features 6x smaller embeddings, global language support, and is now available on Amazon Bedrock. This addresses the fact that 80% of the world's data is in video format but rarely used in AI models. Check out Marengo 3

• Slack's Vision AI for Team Collaboration – Slack brings together people, data, apps, agents, AI and workflows in a single platform that provides the context needed to boost productivity at AI speed. Discover Slack's Vision AI

• GPT-5.1 Codex Max Released – GPT-5.1-Codex-Max is a new frontier agentic coding model available in Codex, built on an updated foundational reasoning model to deliver faster, more intelligent and more token-efficient support across the development cycle. Explore the announcement

• DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models – DeepSeek’s V3.2 release delivers gold-level mathematical and complex-reasoning performance through its maxed-out V3.2-Speciale variant, matches GPT-5-level results on multiple benchmarks, and intensifies competitive pressure on OpenAI and Google. Read the research

• Anthropic's Claude Opus 4.5 with Sync 5.4.7 –Anthropic announced Claude Opus 4.5 with Sync 5.4.7, representing significant improvements in model capabilities and synchronisation features for enterprise applications. Explore Claude Opus 4.5

• Amazon is introducing updates to its main line of AI models – Amazon is introducing updates to its main line of AI models, including Nova Forge, a generative AI model designed to let sophisticated users grab versions of Amazon’s Nova models and customise them with their own data. Discover the details

Links We are Loving

• KPMG: What it takes to deliver AI value — KPMG's latest analysis examines what organisations actually need to deliver AI value, moving beyond implementation to sustainable results and measurable returns on investment.

• None of them are saying ASI is a fantasy, or that it's probably 100+ years away— Leading AI researchers aren't saying ASI is fantasy or 100+ years away, sparking renewed debate about artificial superintelligence timelines and development trajectories.

• Stephen Klein on market dynamics — Stephen Klein Premium: "The Official 'Bubble' Post To Help Anyone Who Looks Back At The Carnage And Ask In Wonder, 'How Did That Happen?"The upcoming crash is explained.

• Learn to surf before riding the AI wave — AI certainly isn’t going anywhere, but are we running before we can walk? And is today’s tech hype aligned to the day-to-day reality of our industry?

• How AI and automation are streamlining material buying — The construction industry is experiencing a major shift in material purchasing as AI and automation move beyond buzzwords to actively transform how subcontractors source, compare, and buy materials.

• Call for built environment AI strategy —Autodesk and the Association for Consultancy and Engineering (ACE) have called for the government to create a national AI in engineering and built environment strategy.

• One of Google’s biggest AI advantages is what it already knows about you — A Google Search exec said that one of the company’s biggest opportunities in AI lies in its ability to get to know the user better and personalise its responses.

• College Students Flock to a New Major: A.I.  — At M.I.T. a new program called “artificial intelligence and decision-making” is now the second-most-popular undergraduate major.

• OpenAI takes an ownership stake in Thrive Holdings to accelerate enterprise AI adoption — OpenAI and Thrive Holdings will collaborate to deliver direct, scalable impact across core enterprise operations, initially focusing on accounting and IT services where high-volume, rules-based, and workflow-intensive processes allow OpenAI’s platform to generate immediate benefits.

• I’m an A.I. Developer. Here’s How I’m Raising My Son — The New York Times opinion piece on how parents can prepare children for an AI-transformed future, focusing on skills that will remain valuable.

• Internal AI adoption accelerates across tech sector— Technology companies are increasingly using their own artificial intelligence models as they seek to demonstrate the transformative potential of this fast-evolving field to would-be buyers.

• AI alone won’t build construction resilience — Artificial intelligence can play a vital role in construction but it cannot replace strong planning and human expertise, writes Nick Gray, chief operating officer for the UK and Europe at Currie & Brown.

• Anthropic hires lawyers as it preps for IPO — Anthropic is prepping for an IPO that could come as early as 2026, and it has brought on law firm Wilson Sonsini to help kick off the process, and the company is tackling an internal checklist to prepare it for what could be one of the largest IPOs ever.

• Digital Construction Today: Accelerating the Future Report - Comprehensive report on accelerating the future of digital construction practices and AI integration across the built environment sector.

• OpenAI Infrastructure Partners Rack Up $96B in Debt for Mega Data Centers — Companies supplying data centres, chips, and “compute” processing power to OpenAI have taken on about $96 billion in debt to fund their operations, according to an analysis by the Financial Times.

• Google DeepMind's Weather Prediction AI — WeatherNext 2 is the state-of-the-art family of weather forecasting models from Google DeepMind and Google Research.

• HSBC calls out OpenAI— OpenAI won’t make money by 2030 and still needs to come up with another $207 billion to power its growth plans, HSBC estimates.

• Marriott brings AI concierge to Napa, transforming how guests pick wine — The Napa Valley Marriott has become the first destination hotel in the region to embed Preferabli’s patented AI technology into its guest experience, offering personalised food, wine and spirits recommendations at scale.

The Spotlight Podcast

The Consulting Goldmine: How Productisation Could Unleash a Billion-Dollar Market

What if everything consultants believe about AI's threat to their industry is backwards? Ikum Kandola, co-founder of The AX.ai, makes a compelling case that artificial intelligence is actually unlocking opportunities so vast the consulting market could expand a thousandfold. The catch?

It requires a fundamental shift in how consultants think about their own value. Kandola argues that once consulting services become productised through AI, delivery times shrink from months to weeks and costs drop dramatically. A consultancy serving 50 clients annually on bespoke engagements could suddenly serve 5,000 through productised offerings. The consulting market isn't constrained by demand but by delivery speed and price.

Listen Now

Listen Now or watch now

Read the blog

Event of the Week

AI-2025 Forty-fifth SGAI International Conference on Artificial Intelligence

16-18 DECEMBER 2025 | Peterhouse College, Cambridge, UK,

The conference is aimed at those who wish to update themselves with news and views of recent developments, understand how other groups are applying the technology and exchange ideas with leading international experts in the field. Our goal, as always, is for our annual conference to be a meeting place for the international artificial intelligence community. The two-track conference and the workshops are designed to provide a wide range of options for delegates, whether they are newcomers to the technology or seasoned practitioners.

The social programme provides an opportunity to relax, meet old friends and make new contacts. The scope of the conference includes the whole range of AI technologies and application areas. Its principal aims are to review recent technical advances in AI technologies and to show how this leading edge technology has been applied to solve business problems. Register now

One more thing

That’s it for today!

See you soon,

James, Yoshi and Aaron—Project Flux