Joined March 2025
- Tweets 649
- Following 26
- Followers 2,724
- Likes 818
191 Photos and videos
Pinned Tweet
16 Sep 2025
After only 4 months, we are delighted to announce our acceptance into the @nvidia Inception program, the elite accelerator for start ups driving breakthroughs in AI.
This validates our focus: verified, ethical, human-powered datasets enabling the of AI. The trusted fuel for AGI is here - join the revolution!
#DGN #AI #NVIDIAInception #Data
16 Sep 2025
37
40
106
25,057
Mar 6
The shift to agentic AI is accelerating.
@OpenAI's latest release pushes further into systems that can plan and execute tasks autonomously.
When this is applied to business, enterprise advantage won’t come from agents or models alone. It will come from who can apply sovereign data best in
- proprietary training loops
- protected datasets
- provable provenance and legal defensibility
When done correctly, this transition will enable huge growth opportunities.
theverge.com/ai-artificial-i…
5
9
617
Mar 5
As the USA's focus on Iran continues, @AnthropicAI - reportedly used in planning the recent strikes - refused to change their approach and remove safeguards around AI use in autonomous weapons and surveillance.
The response from Washington:
The company was labelled a “supply chain risk” by the Pentagon and Secretary Pete Hegseth.
Anthropic says it will challenge the move in court. So what did this show?
bbc.co.uk/news/articles/c3rz…
1
3
7
421
Mar 5
Beyond the politics, this highlights a deeper technical issue.
Today’s AI models still struggle with critical real-world issues:
1. Real-world context
2. Adversarial reasoning
3. Evolving and ever-changing environments
In high-stakes systems, data quality and human oversight still matter more than model scale.
1
3
3
228
Mar 5
What's more, @OpenAI jumping into take on the deal instead has had adversarial effect.
It triggered a wave of backlash against Altman and OpenAI, even pushing Claude above ChatGPT to #1 on the App Store.
While the real issue may be capability, what's happening now is a debate framed around ethics vs policy.
3
6
205
Mar 2
The AI market has seen big deals announced in the past week.
@Meta with @AMD and @Google for GPUs and TPUs
@nvidia with both @Lumentum and @CoherentCorp.
@OpenAI's funding round.
@AnthropicAI and @OpenAI's tussle with the US Government.
Within this, there is a growing focus on national AI stacks - sovereign compute, localized infrastructure and state-backed AI programs.
2
4
6
263
Mar 2
Here’s the shift 👇
Sovereign compute works best if the data pipelines are sovereign too.
That requires traceable or localized training.
Auditable datasets is a must have.
Traceable provenance becomes critical.
You also need controlled access environments.
Compute Data = Competitive advantage.
That’s exactly why we built D-GN around human-verified datasets, on-chain proof layers and enterprise-ready data governance.
Sovereign AI isn’t just a chip story. It needs to be a full-stack data strategy.
3
5
160
Feb 26
Reuters reports that ECB President Christine Lagarde sees no wave of AI-led layoffs - yet.
AI isn't causing mass unemployment yet. But it isn't creating much for structured, accessible job funnels at scale either.
That's the gap.
If AI is reshaping labour, people need an on-ramp.
D-GN wants to help build the ramp.
✅ Human-led data
✅ Real AI learnings
✅ Verifiable contribution records
✅ Earn while you learn pathways
AI shouldn't just optimised companies. It should expand participation.
At D-GN, we're going to bring people into the loop. Properly.
reuters.com/business/ecb-see…
#AI #FutureofWork #DataEconomy
1
3
5
173
Feb 25
Most production #AI systems fail at the margins.
It's rarely because the architecture is weak - but because the training data is generic.
In competitive industries, a data difference compounds. 🧠📉
2
2
9
217
Feb 25
Base model weights could start out similar - but every update will have an outsize advantage.
In highly competitive industries, a 1–2% performance increase can mean:
⚡ Huge energy optimisation gains
🏭 Fewer defects and edge cases
💻 Faster debugging cycles
🛡️ Reduced compliance risk
1
1
2
82
Feb 25
The real moat in 2026 isn’t the model.
It’s the sovereign, traceable, human-validated data pipeline behind it.
In frontier markets, sovereign data becomes the critical advantage - that's the strategic infrastructure. 🏗️
1
2
81
Feb 23
'Big Tech' is projected to spend a wild $650 billion on AI computing in 2026.
That's nearly triple the spend in 2024. It's infrastructure-level commitment.
The AI race is no longer about models alone. It's about
➡️ Compute scale
➡️ Data Centre dominance
➡️ Energy Access
➡️ Proprietary Training Data
Each level is a competitive advantage for huge AI progress, not just GPUs.
bloomberg.com/news/articles/…
1
2
5
136
Feb 23
Enterprises that treat data provenance, auditability and human-led training as strategic advantage will compound fastest.
AI is consolidating at the bottom (chips hyperscale).
Enterprises need to take the opportunity in the middle - data, governance and application.
Sign up to the D-GN waitlist to learn more about how you can maximise competitive advantage and own your data layer: d-gn.io/
2
4
102
Feb 20
The UK-led international AI safety coalition has just got two new joiners - @OpenAI and @Microsoft.
They join governments, frontier labs and academic institutions backing the UK AI Security Institute’s global alignment effort.
This is significant for AI. The UK is positioning itself as a convenor of AI safety standards. Not just a regulator, but a coordinator of international safeguards.
As AI capability accelerates, influence will sit with those shaping evaluation frameworks, alignment benchmarks and safety testing protocols
These guardrails - who sets them and how companies can prove to be within them - is going to be critical.
🔗 gov.uk/government/news/opena…
2
4
127
Feb 19
#AI performance was never just a model architecture problem.
It's a data pipeline problem.
We've seen this firsthand building annotation infrastructures:
1⃣How data is sourced changes what a model can generalise
2⃣How it's annotated helps shape decision boundaries
3⃣How it's verified determines failure modes
4⃣How it evolves across cycles defines shelf life
What does this mean? Human-led annotation isn't optional - it's structural.
1
4
9
230
Feb 19
There’s also a second-order effect:
When you bring people into the AI training loop, you help build distributed AI literacy - a critical world focus.
We also want to surface real-world usage patterns, create feedback grounded in lived experience and expand to a workforce that understands how models fail.
So why would you train AI 'on' people? It should be trained 'with' them.
1
1
2
75
Feb 19
The next wave of competitive advantage won’t come from bigger models alone.
Unique, sovereign datasets will become critical.
They add true differentiation when combined with custom reinforcement cycles, human-aligned validation and transparent provenance.
The extra benefit is the communities that get trained alongside the AI itself.
Architecture scales. But alignment is engineered. At D-GN, we working to align the architecture with the world.
3
4
89