Web Summit 2024: Bet More on AI

So yeh. I attended Web Summit 2024 in Lisbon, Portugal—Europe’s largest tech summit bringing together startups, investors, and technology leaders.

Summary

"Bet more" (on AI) is my key takeaway from this event.

I was left with the clear feeling we are not being nearly aggressive or imaginative enough when it comes to AI. The standout moment was witnessing Alibaba’s Accio.com demo—a glimpse into where B2B commerce and AI-powered agents are heading.

Key Themes:

• AI as the dominant theme across sessions
• Small, specialized AI models gaining traction over monolithic systems
• Service agents and agentic AI architectures emerging
• Community-based marketing (small private vs large public social media)
• Increasing focus on specific niches, audiences, use cases
• Strong representation from crypto, Web3, payments & fintech
• Robotics + AI convergence predicted as major 2025 trend

Monday: Opening Sessions

Portugal’s AI Strategy

The Prime Minister of Portugal pledged a Portuguese LLM to preserve PT language and culture, plus an AI education assistant for school kids.

The Mayor of Lisbon gave a thoughtful talk about the need to give back from wealth to social causes, positioning social care as the sustainable basis for ongoing innovation. The ongoing strikes (schools, and on the final day the entire metro system) highlighted real social challenges beneath the tech optimism.

Twelve: Carbon Transformation

Speaker: Etosha Cave, Co-Founder

Twelve transforms CO2 into liquid hydrocarbons—they call it "carbon transformation."

Technology: The Opus System uses a chemical/physical process (as opposed to biological processes from competitors like Lanzatech).

Scale: Just raised $650M (equity + debt + project financing) to scale to 100MWs—equivalent to 100,000 cross-country passenger jet flights per year.

I had breakfast with Etosha at the speaker lounge—we discussed hydrogen, clean tech, crypto, regenerative finance, and DeSci (decentralized science). Fascinating intersection of climate tech and Web3 financing models.

AI Safety Debate

A somewhat adversarial panel featuring different AI safety philosophies:

Radical Openness: Thomas (representing open-source AI approach) believes the best way to ensure AI safety is radical transparency and open development.

Safety Standards: Max argued that AI (particularly AGI) should be subject to safety standards similar to drugs, cars, and other regulated tools.

Agreement: Both agreed AI would unlock robotics, with Thomas predicting this becomes a major trend in 2025. Small models that are fast and constrained enough for physical contexts will finally animate the robotics components we’ve had engineering capability to produce for years.

Pharrell Williams: Creativity and Culture

Really enjoyed this despite poor interviewing. Some of Pharrell’s insights:

"The human dream should be about spending as much time as possible doing something that you love."

"Strong faith and a tablespoon of delusion lets you taste and feel your vision. It’s what gives you conviction."

"You are the culture. Look in the mirror, the fire is in you."

"Each one of you has a responsibility to give back to that which gave to you."

Tuesday: AI Models and Applications

Small AI Models and Adapters

Speaker: Hugging Face representative

Language models have different sizes based on parameter count. Small models can do a lot!

Training Approach: Small models are pre-trained similarly to large models but for longer (67-200x) on more focused datasets. This reduces parameters by pruning nuanced routes and leaving only clearer pathways.

Advantages:

• Speed (less compute required)
• Privacy (can run locally)
• Portability (deploy to edge devices, robots)
• Even our fridges will have embedded small models

The Spectrum of Open Source:

• open weights – model weights publicly available
• open data – training data publicly available
• Full open source = both weights and data

Modular Architecture:

• Use adapters as "hats" that transform a small base model
• Same ~1GB model can become a summarizer, translator, or domain expert
• Derivative fine-tuning creates evolutionary families of models for increasingly specific needs

Synthetic Data: "We’ve run out of organic data on the internet." Solution: Use AI to generate synthetic high-quality datasets for specific domains (mathematics, industry-specific knowledge, etc.).

Alibaba: Accio.com B2B Commerce

This was the standout talk of the conference. I went from 60% attention to 100% attention.

Accio.com offers:

• B2B search
• B2B wiki—every SKU gets a dynamically updating page
• Agent—chatbot interface as "your personal sourcing agent"

What Made It Remarkable: Watching Accio create and maintain wiki pages for products, push out notifications made the pages feel alive—actionable, social, and relevant.

The Unsettling Contrast: The business of creating pages felt trivial—like an assumption or afterthought. The real magic was the content within the pages: dynamic, updating, connected.

This feels like a canary for where commerce platforms are heading. The page itself becomes commodity; the intelligence, connections, and real-time data become the value.

Microsoft: AI as General Purpose Technology

Speaker: Senior Microsoft executive

Framed AI from the economic perspective of General Purpose Technologies (GPT).

Historical GPTs:

• Electricity
• Internal combustion engine
• Computing
• Internet

All industrial revolutions have been driven by a new GPT. Each GPT creates its own tech stack.

Electricity’s Economy:

• Fuels
• Power generation
• Electric grids for distribution
• Transformers and circuit breakers
• Wiring, switches, sockets
• Appliances
• Manufacturers and users

Electricity created appliances that made people’s lives easier. Companies like AEG licensed Edison’s tech and helped build national economies.

The Challenge: What will the components of the new AI economy be? What are the "appliances" of AI?

Wednesday: Social, Media, Design

Work-Life Integration

Speaker: Lattice (HR platform)

• Importance of being 100% at work OR at home—clear boundaries and intentional management
• Lattice connects people’s work with corporate purpose

Trust and Social Media

Panel: Social media leaders discussing content moderation, misinformation, platform responsibility

Quote: "Creating a pipeline of information to all the media platforms" – Joe Benarroch

Observations:

• No easy answers to the social media challenge
• Relies on trust, checks/balances, and rapid response
• AI and deep fakes did NOT overrun our ability to discern truth this election cycle (surprising)
• None mentioned my view that the only solution to ‘truth’ on the internet is economic

Rather than enshrining an ideal like "free speech" which we agree is good, certain behaviours we agree to be malicious are made prohibitively expensive. They are not disallowed, just economically unsustainable, in much the same way that prevention is better than any cure.

From Kernel’s Module on Money & Speech

AI and Creativity

Speakers: Picsart and others

• Comparison to impact of Photoshop—ultimately additive to creative space
• "Democratizing creativity" as mission
• Referenced "Greatness cannot be planned" concept—exploration over optimization

Thursday: Service Agents

Agentic AI Architecture

This session crystallized the future of AI applications.

Service Agents: Specialized AI agents for specific business functions.

Prediction: The future might be lightweight SaaS with AI-mediated integrations. Lots of discussion about Salesforce’s potential disruption.

Pattern: Lots of small/distilled models > one huge model

Architecture:

• One "point of contact" agent orchestrates the others
• Back-end agents (API-driven, structured, straightforward)
• Front-end agents (interface with people, handle ambiguity)

Economic Impact: Tony predicts → 90% reduction in labor → means either reducing headcount, 10xing output, or underperforming.

Implication: If you have the right idea and economics, AI will help you scale very fast.

Challenge: Productization remains critical. Example: Google is engineering-led, which is why Gemini and Google Workspace lag behind Gmail and GCP. Emphasizes the importance of product thinking over pure engineering.

People I Met

• Alissa Leeds – Ex-Meta, NFL. Looker consultant, marketing & data analytics. Really smart, we hung out throughout the conference.

• Brendan Yell from FusionAuth – Authentication platform, aiming for post-event conversation about identity and auth trends.

• Etosha Cave from Twelve – Carbon transformation company (mentioned earlier).

• Alex Bridges from Wavecast – Hybrid product/services business for livestreaming experiences. Early sites built on WordPress but migrated to Laravel due to multisite complexity.

• Dr. Dominik Schürmann from heylogin – B2B password management solution using OAuth & device-based MFA.

• Tim Deeson – Agency founder I know from past, sorted me out with a Speaker guest pass upgrade. Would make an excellent startup/agency coach.

• Andy Budd – Founder of Clearleft, now VC with Seedcamp. Continues coaching, writing, speaking. Another excellent potential coach for founders.

• Yoni Belousov from .link – Discussed crypto, open web identity challenges, making identity self-sovereign (users own domains rather than platform usernames) while still benefiting from social connection and distribution.

• Zach Seward from NYT – Discussed trust and information markets, how to achieve trust in today’s web through economic mechanisms rather than pure moderation.

• Leslie Keeler Saglio – Coach specializing in women + AI.

Expo: Startups and Emerging Tech

Outside scheduled talks, I spent hours walking the expo pavilions. Loads of startups! The stalls changed daily so I probably only caught 10-15% of exhibitors.

Notable Startups and Trends

News & Content:

• Newscatcher API – Clean, enriched, ready-to-use news data

Travel / Adventure (very specific niche focus):

• Sabię – TravelTech combining social connection, sustainable choices, AI planning
• Explorer Elite – Adventure community connecting travelers
• Ambitio – Platform for travelers and dreamers

Crypto / Web3 / Fintech / Auth / Identity:

• Banxe – Banking with money and crypto
• Reputy – Showcase skills through real-world experiences using Soul-Bound Tokens (non-transferable NFTs) as modern CV
• RedStone Finance – Fastest-growing Oracle for DeFi, available on 50+ chains. Interesting: they use Arweave for data archiving.
• SpacesDAO – Artists, architects, developers designing tailored metaverse experiences
• Friendo.cash – Innovative Money Affiliated Network connecting financial and social networks
• Social.plus – Pre-built social and community features for apps with AI-powered insights
• Beam – "OnlyFans for brainiacs" helping experts monetize knowledge via live video
• Modatta – Monetize your data privately, brands pay you directly to interact
• Rewardy.gg – Embedded rewards system for web3 apps
• Aeter Edge – PSP for high-risk industries with clever payment-routing IP
• AZKR – "League of Identity" allowing crypto assets sent to emails and social media accounts ("airdrop by email")
• Moja – USDC-powered financial services for families, friends, communities across borders
• Cryptr – Auth features like SSO for enterprise readiness
• Forkbomb Solutions – Multiplatform cryptographic stack for EUDI-ARF identity, digital product passport, post-quantum cryptography
• AMLBot – Automates AML/KYC procedures, reduces compliance expenses
• Vexor.gg – AI-driven reputation protocol for gaming user acquisition, retention, engagement
• COCA Wallet – MPC security, non-custodial debit card, no KYC
• Trustee Global – Buy crypto with bank card, instant exchange, withdraw to bank card
• Odys Music – Stream music, connect with friends, own tokens. Fans buy tokens monthly, funds distributed based on listening share.
• Peachz – Like Bandcamp but Web3, built on Solana
• RNS.id – World’s first government-issued blockchain identity

AI-Powered Development:

• Locofy.ai – Frontend development at lightning speed, design to code
• Builder.ai – "Building an app so easy, anyone can do it"

Other Notable:

• Viads – Maximize revenue & brand awareness via premium ad networks
• Base.com – Unite all sales channels, manage from one place
• Mycrobez.ch – Mycelium composites from 100% Swiss organic waste and mushroom roots, scalable alternative to plastic foam
• Niantic Labs – Spatial intelligence platform transforming how people and machines understand the physical world

AI-Powered Investing:

• Unitor.ai – AI financial assistant analyzing data, setting goals, guiding to financial freedom
• ACM Investments – Help clients make informed investment decisions
• Reinvent – Combine human expertise and AI for easier money management

Key Takeaways and Reflections

AI: Not Aggressive or Imaginative Enough

What would I want AIs to do for me? → CLEAR THINKING. Then AI can help execute and scale.

Questions to ask:

• What expertise barriers can AI destroy to open fast-mover opportunities?
• Where does value accrue—the large models in the middle, or the small models and connective applications at the edges?
• Counter-intuitively, AI might see value distribution out to smaller businesses rather than further concentration

Specific opportunities:

• Small models, front/back-end agents, agentic teams
• Personalisation at scale
• Generating synthetic training data
• Destroying expertise barriers to create new markets

AI as General Purpose Technology

AI is the next great GPT (General Purpose Technology) with implications for the next industrial revolution. Like electricity created an entire economy (generation, distribution, transformers, appliances, manufacturers), AI will create its own tech stack.

The question: What are the AI equivalents of light bulbs, motors, and refrigerators?

Overall Event Impressions

Positive:

• Quality of center stage talks was good
• Pleasantly surprised by male/female balance and inclusion of sustainability theme
• Totally full-on event, comparable to SXSW but with more focus on web, startups, investment
• Strong representation from crypto, Web3, payments & fintech

Practical:

• Worth getting a hotel close—Lisbon traffic is a nightmare
• E-bikes and Bolt rides were affordable but time-consuming
• Would attend more side events next time

Strategic:

• Important opportunity to ground truth and contextualize emerging technology trends
• Robotics + Biotech → applying digital expertise to controlling physical world—likely major trend for 2025 and beyond

Looking Back from 2025

Reading these Web Summit 2024 notes just one month later (early December 2024), several predictions are already manifesting:

What’s Accelerating:

• Service agents and agentic AI—multi-agent systems (MAS) becoming mainstream conversation
• Small, specialized models—the "adapter" architecture proving out
• AI + robotics convergence—physical AI emerging as major theme at SXSW 2025
• Economic models for truth—still largely unaddressed by platforms

What Remains Challenging:

• Productization over engineering—still the key differentiator
• Trust and social media—no consensus solutions emerged
• Crypto/Web3 mainstream adoption—still mostly speculation vs. utility

Accio’s Impact: The Alibaba B2B commerce demo remains the most memorable moment. The idea that page creation becomes trivial while dynamic content becomes everything feels increasingly prophetic. Content management systems may need to evolve from "managing pages" to "orchestrating living, connected information."

The "bet more on AI" thesis strengthens daily. The question isn’t whether AI transforms industries, but whether organizations move fast enough to capture the opportunity.

---

Event: Web Summit 2024
Location: Lisbon, Portugal
Dates: November 11-14, 2024
Published: November 18, 2024

Related Content:

• SXSW 2025: AI, Physical Intelligence, and The Beyond – Follow-up conference with physical AI themes
• SXSW 2024: Technology Supercycles and AI Transformation – Earlier in 2024
• Algarve Tech Summit 2025 – Local tech community building

Source: Originally published November 2024 as internal conference notes, sanitized and enriched for public release December 2025.