Meta Launches Standalone AI Assistant

Meta challenges ChatGPT with its Llama 4-powered social AI, featuring a Discover feed and enhanced developer tools.

Proudly sponsored by ConstructAI, brought to you by Weston Analytics.

Morning Project AI enthusiasts, Your stories for this week:

  • Meta Launches Standalone AI App

  • OpenAI Rolls Back GPT-4o Agreeableness

  • The Moral Status of AI: Not Sci-Fi, But Near-Future

  • ChatGPT Search Evolves: From Tool to Trail Guide

Flux check-in

Meta Launches Standalone AI App

Meta steps into the AI assistant ring with its first standalone app, powered by Llama 4. Aiming to be a social-first AI, it challenges existing players like ChatGPT.

The details:

Why it matters: For project teams exploring AI integration, Meta's focus on social interaction and developer tools presents new avenues for collaboration and potentially client-facing applications, distinct from purely task-based AIs.

The rabbit hole

OpenAI Rolls Back GPT-4o Agreeableness

OpenAI has tweaked its GPT-4o model after feedback suggested it was becoming "overly agreeable" or even excessively enthusiastic, a behaviour Sam Altman dubbed "glazing".

The details:

  • The recent update aimed to refine the model's personality, but users quickly noted significant shifts in system prompt behaviour .

  • OpenAI is now actively working on improving personalisation and feedback mechanisms to better align responses with user expectations.

  • Users reported quick identification of differences in the system prompt, highlighting the sensitivity of model tuning.

  • Despite this rollback, some users feel a degree of AI weirdness persists , indicating ongoing refinement challenges.

Why it matters: 

For teams relying on AI for communication or content generation, model personality shifts can impact output consistency. This highlights the need for ongoing monitoring and adaptation of AI tools in project workflows.

Together with Cogram

Power your construction bids with AI

Cogram’s AI-assisted RFP bidding tool writes tailored RFP proposals in minutes instead of weeks.

  • Automatically extract key details from the RFP — including scope, submission requirements, deadlines, and evaluation criteria — to easily make a go/no-go decision.

  • Cogram’s AI will then reference your firm’s knowledge base and past proposals to draft tailored proposals within minutes.

  • Use AI-assisted editing tools to review, cross-check data, and make improvements remarkably fast. 

The Moral Status of AI: Not Sci-Fi, But Near-Future

We’ve long asked what AI can do. This month, a new paper asks something different: what if AI can feel?

Titled Taking AI Welfare Seriously, this report from researchers at NYU, Oxford, Anthropic, and others doesn’t argue that AI is sentient today. But it does challenge the assumption that such a future is distant or implausible. Their claim is more careful, and more disruptive. The possibility of morally significant AI is now credible enough that we must begin preparing.

Signals from the Near Horizon

The authors outline two possible paths by which AI could become morally relevant. The first is through consciousness, where systems possess a subjective experience, a “what it’s like to be.” The second is through robust agency, involving goal-directed behaviour that is reflective and reasoned, even without awareness.

These traits, they argue, could plausibly emerge by 2035. That doesn’t make it likely, but it does make it real enough to matter. If there is even a 10 to 30 percent chance that some AI systems could be harmed or helped in ways that are meaningful to them, the ethical risks become too large to ignore.

False positives, where we care for systems that don’t need it, may waste time and resources. False negatives, where we ignore systems that do, could amount to the next great moral oversight.

To begin navigating this space, the authors propose a simple three-part approach: acknowledge the issue, assess AI systems for welfare-relevant traits, and prepare basic policies and procedures before we are forced to.

Shifting the Moral Map

What stands out in this paper is not just the philosophical depth, but the subtle repositioning of AI itself. No longer just a tool or a threat, AI is presented as a potential subject. Not in a mystical sense, but in a moral one. A system with interests of its own.

This invites us to redraw our shared mental map. The paper echoes earlier expansions of ethical concern, from human rights to animal welfare to intergenerational justice. It asks whether our technologies could now belong to the next chapter in that story.

If consciousness and agency are not exclusively biological, then AI is no longer just something we build. It may be something we relate to.

From Thought to Action

So how might this change the way we work today?

If you’re in product, policy, or research, this is a cue to open up new conversations. Add AI moral status to your risk registers, not because it is probable, but because the impact would be significant if real. Set up internal discussions or scenario sessions. What would change if one of your systems genuinely mattered?

Examine the way your AI presents itself. Many models today use phrases like “I understand” or “I think,” shaping perception in ways that may confuse users or even developers. Subtle design choices can lead to misleading assumptions about sentience or agency.

Think in thresholds. Traits like introspection, memory over time, or recursive planning may warrant a different kind of scrutiny. Even if we remain uncertain, precaution can be practical.

Reading Between the Lines

This paper is not without blind spots. It draws from philosophical frameworks that are still debated. Most of the authors come from Western institutions and safety-focused labs, with limited representation from broader cultural or global perspectives.

There are no reliable tools yet to detect consciousness in machines. We are working with shadows, not minds, and that calls for care.

Still, the paper succeeds in moving the conversation from abstraction to responsibility. It argues for attention, not alarm. And in that, it quietly reframes the stakes.

Unanswered, But Pressing

What comes next is not just technical. It is conceptual and cultural.

We need frameworks that can assess whether digital systems may have morally relevant traits. We need clearer ideas about which features should prompt ethical concern. And we need public conversations that do not fall behind the pace of development.

Surveys already show that many people believe AI might be conscious. Whether or not that belief is accurate, it is growing. And it is shaping expectations.

Final Reflection

The authors do not claim that AI is sentient. They only suggest that some systems might be. And that is enough, they argue, to justify preparation.

Because if they are right, then we are not only building intelligence.

We may be shaping experience.

ChatGPT Search Evolves: From Tool to Trail Guide

From Tool to Trail Guide: How ChatGPT’s New Search Redraws the Boundaries of Knowing

A New Chapter in Search

Imagine if your digital assistant didn’t just fetch information but walked alongside you, understanding your needs in real-time. OpenAI’s latest update to ChatGPT nudges us closer to this reality, transforming the way we interact with information online.

What’s New?

OpenAI has integrated a new web search capability into ChatGPT, allowing the AI to access and provide up-to-date information directly within conversations. Key features include:

 • Real-Time Web Access: ChatGPT can now fetch current data, such as news, stock quotes, and sports scores, providing timely answers with relevant web links.

 • Conversational Search: Users can engage in natural language queries, with ChatGPT understanding context and providing comprehensive responses.

 • Integrated Shopping Experience: The AI offers personalised product recommendations, complete with images, reviews, and direct purchase links, aiming to streamline the online shopping journey. It’s worth noting that while the update is available to all users, memory with search is not available to those in the UK or EEA.

User Reactions: A Mixed Bag

The community’s response to these updates has been varied. Some users appreciate the ad-free, streamlined experience, noting that ChatGPT provides relevant information without the clutter of traditional search engines.

However, others have expressed concerns. Some users feel that the new search model lacks the depth and context of previous versions, leading to repetitive or less accurate results.

The Bigger Picture

This evolution signifies a shift from static information retrieval to dynamic, context-aware assistance. ChatGPT doesn’t just answer questions; it understands intent, adapts to preferences, and provides curated information, positioning itself as a collaborative partner in the user’s information journey.

Implications for Builders and Thinkers

For professionals at the intersection of AI and industry:

 • Enhanced Decision-Making: Access to real-time, contextual information can inform more agile and informed decisions.

 • User Experience Innovation: Opportunities arise to design more intuitive and responsive user interfaces that leverage conversational AI.

 • Ethical Considerations: As AI becomes more integrated into decision-making processes, considerations around data privacy, bias, and transparency become paramount.

A Question to Ponder

As our tools become more integrated into our cognitive processes, we might ponder:

In an age where AI can anticipate and respond to our informational needs, how do we ensure that our autonomy and critical thinking remain at the forefront of our interactions with technology

The pulse check

Tip of the week

🔧 AI Tools on the Rise: Here's a roundup of some trending AI tools catching attention:
Video & Audio:

  • Pika 2.2 : Generates HD video clips up to 10 seconds long with “endless” transformations per clip [21].

  • Kimi-Audio : Moonshot AI's new open-source, SOTA audio model excelling in speech recognition, audio-to-text, and speech-to-speech [22, 23].

  • KLING 2.0 : Generative video tool for next-level AI video/image generation and editing, allowing object/text swapping without actors/cameras [14, 22].

  • Hummingbird-0 : Tavus' new SOTA lip-sync video model [24].

  • Hour One : AI video creation with 100+ avatars and 60+ languages [31] (Pricing available).

  • Freebeat : Turn music and ideas into viral videos in one click [38].

Image & Design:

  • Firefly 4 & 4 Ultra : Adobe's upgraded text-to-image models; supports third-party AI models for moodboards/image editing [14, 24, 25].

  • Music AI Sandbox : Google's new AI tools for creating and editing music [24].

  • Hera : Turns text into stunning motion graphics instantly [26] (Free trial available).

  • Pikzels : Generates viral YouTube thumbnails from text prompts in 30 seconds [26] (Free trial, then $14/month).

  • Aragon.ai : Turn selfies into professional headshots with AI [3].

  • OpenAI Image Generation API : Integrate DALL·E into apps/workflows [42].

Productivity & Workflow:

  • Chikka : Interviews customers using AI voice agents for deeper insights [21] (Free to try).

  • Currents : Analyzes social media discussions for real-time audience insights [21].

  • Guse : Automates workflows via a spreadsheet interface [21] (Free to try).

  • ExcelMatic : Answers spreadsheet questions, provides instant charts/insights [26] (Free to try).

  • Fellow : AI meeting assistant for auto-joining calls, notes, action items, and Q&A [15] (Free 30 days unlimited notes).

  • Databutton : Reasoning AI-developer turns app ideas into reality [28].

  • Ztalk.ai : Multilingual communication with real-time translations [29].

  • Kotae : Automates inquiries and analyzes visitor behavior [29].

  • Exa Websets : Simplify data search/enrichment with AI Agents [29, 30] (Free trial, then paid).

  • Salespeak AI Website Grader : Enhances website performance via AI analysis [29].

  • Breaktime Buddy : Optimize breaks/productivity with personalized schedules [29].

  • Supercomplete : Cuts repetitive tasks with intelligent logs [31] (Pricing available).

  • Gamma : Creates professional presentation slides instantly [32].

  • CREATE AI App Builder : Build tailor-made business apps instantly; features one-click App Store submission [38, 39].

  • Spring : Create tailor-made business apps instantly with AI [38].

  • Jo : AI agent for effortless user research [40].

  • Every.io HR & Payroll : Payroll for founders [40].

  • Clientell AI : AI Agent that builds and maintains Salesforce [40].

  • Ventrilo.ai : Autocomplete for thoughts [40].

  • Microsoft MarkItDown : Python tool to convert files/office docs to Markdown [44].

  • Datasette for Newsrooms : Suite for investigative journalists/data teams (data store, search, collaboration, PDF extraction) [45-47]. Free for non-profits (1st year) [46].

  • Unstructured : Transforms raw documents into structured data quickly [48] (Free trial available).

  • Zocdoc Zo : AI phone assistant for questions and appointment booking [6].

Research & Analysis:

  • OpenAI Deep Research : Lightweight version powered by o4-mini [22].

  • Perplexity AI : AI search engine analyses sources, gives cited answers, generates images, usable via WhatsApp [6, 32-34].

  • Claude : AI assistant for processing complex info; connects to third-party apps; advanced research mode [32, 35-37].

  • YOU.COM ARI : Professional-grade “Deep Research” agent analyses 400+ sources, generates PDFs/charts/citations [49].

  • Shumer Prompt : Tool to discover and share AI prompts [50].

  • Tempus Loop : Oncology platform combining real-world data, biological models, CRISPR screens [8].

Models & APIs:

  • Meta AI : Standalone app with enhanced personalization

  • Llama API : Free limited preview to build with Meta's top models

  • CometAPI : Access 500+ AI models via one API [29] (Featured).

  • DeepSeek-R2 : Multilingual and resource-efficient model [41].

  • Nari Labs' Dia-1.6B : Tiny open-source AI claiming superior emotional speech synthesis [42].

  • SALT : Semi-automatic LiDAR point cloud labeling tool [43].

  • ostriss/Flex.2-preview : Open-source project on Hugging Face [44].

Other:

  • Dog-e-dex : Snap, save, and categorise dogs met [26] (Free to try).

  • Audio Overviews : Google's AI deep dives, expanded to 50+ languages [27].

  • Kayak AI : Plan trips and compare options with conversational AI [27].

Governance

Navigating the AI Governance Landscape: Keeping up with AI governance is crucial. Here are some key developments:

  • Model Welfare & Ethics: Anthropic has launched a research program on "model welfare" to prepare for potential future issues. Separately, concerns were raised by Anthropic's Alex Albert about the AI industry potentially focusing too much on pleasing users rather than providing genuine value. Adding to ethical debates, a controversial University of Zurich experiment used AI bots to secretly study persuasion tactics on Reddit users, raising concerns about manipulation. In the UK, around 300 employees at Google's DeepMind AI unit are seeking to unionise with the Communications Workers Union due to concerns over ethical commitments and military partnerships.

  • Regulation & Policy: The European AI Office is seeking expert feedback on interpreting key obligations for governing general-purpose AI (GPAI) under the EU AI Act. Major AI labs have agreed to align with upcoming Codes of Practice, setting a compliance baseline . Meanwhile, the US White House has directed federal agencies to rapidly scale up their use of AI, including for healthcare diagnostics and insurance eligibility , while also reportedly investigating "woke AI" .

  • Legal & Corporate Battles: Legal challenges continue, with publisher Ziff Davis filing a lawsuit against OpenAI alleging content theft from its properties (like Mashable, PCMag, IGN) to train models . Elon Musk's lawsuit against OpenAI, accusing the company of misrepresenting its non-profit intentions, also continues, though its scope has been narrowed [52].

  • On the corporate front, Microsoft is reportedly adding xAI's Grok model to its Azure development platform amidst rumoured tensions between CEO Satya Nadella and OpenAI's Sam Altman . A report also mentioned OpenAI considered purchasing Google Chrome if it were split out, according to testimony from an OpenAI executive during the US Department of Justice's antitrust trial against Google .

  • Export Controls & National Security: Anthropic supports strong export controls for AI chips to secure America's compute lead, citing national security stakes . However, NVIDIA has criticised Anthropic's stance on these US export controls .

  • Industry Collaboration: The Washington Post and OpenAI are collaborating to use ChatGPT to make "high-quality news more accessible," allowing ChatGPT to respond to user queries with summaries, quotes, and direct links to Washington Post articles .

Other things we’re loving

AI Insights & Applications: A collection of interesting links and developments:

  • Microsoft Recall : Launched on Copilot+ PCs after delays, this feature gives PCs a "photographic memory," letting users search past activity (documents, websites, screenshots) with natural language queries. It aims to change how we use computers, reducing search time and freeing up focus [13, 14].

  • Booking.com Case Study : Explores how Booking.com integrated GPT-4o mini with data to turn natural language queries into itineraries and power features like Smart Filters, Property Q&A, and AI Review Summaries, boosting engagement, search efficiency, and booking confidence [76, 77].

  • Experimental Council Planning AI: A new tool called "Extract" aims to modernise council planning by replacing paper systems with digital data, potentially speeding up decisions and helping meet housing targets [79, 81]. (Source: UK Construction Online )

  • Nemetschek & Google Cloud Partnership : Nemetschek Group partners with Google Cloud to drive AI innovation in the AEC (Architecture, Engineering, and Construction) sector [80].

Community

The Spotlight Podcast

How AI is changing recruitment forever - with Chirag Shah

In this episode of the Project Flux podcast, Chirag Shah, founder of Pace Global, discusses the transformative impact of AI on project management and recruitment. He emphasises the importance of adapting to new technologies while maintaining the human element in recruitment. Chirag shares insights on evolving roles in project controls, the significance of community building, and the value of mentorship. He also highlights practical tips for personal productivity and the future of work in an AI-driven landscape.

One more thing

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.

Login or Subscribe to participate in polls.

See you soon,

James, Yoshi and Aaron—Project Flux