š Agentic Revolution
š Hello! IāmĀ Ivan. I write monthly about the playbooks and hidden tactics of the top 1% of founders and investors. Subscribe to surf emerging tech waves:
Summary
šĀ Agentic Revolution: hype, reality, and whatās coming.
š Pitch Deck GPT: to wrap or not to wrap.
š The reality of PMF: Zuck and Levelsio.
š EU-Inc: help shape the future of EU startups.
šµĀ Iberian Deals: 21 deals in Spain (>ā¬178M).
š« Meme of the month: Amazon remote work.
Want to sponsor Startup Riders? Fill out this form.
This newsletter is proudly sponsored by Yellow Birds Creative.
Creating organic content for your socials is tough. It's even tougher to do so in a way that accurately reflects your brand identity. We offer three packages:
Content creation: photography, video, video podcasting, copywriting.
Social Media Management: taking care of all your content creation needs.
Project Management: ideate, plan, and manage your marketing initiatives.
š Agentic Revolution
āWeāre at the beginning of a new Industrial Revolution. But instead of generating electricity, weāre generating intelligenceā¦ [Open source] activated every single company. Made it possible for every company to be an AI company.ā
ā Jensen Huang
We might be approaching a āChatGPT momentā for AI agents.
It seems like humans might have figured out how to scale intelligence.
Thereās a lot of noise out there.
So hereās are a few key insights and frameworks from the best resources Iāve found to help you wrap your head around this emerging tech wave:
1. AI Agents are probably simultaneously over and under-hyped
"I think the AI boom will rhyme with the dot-com bubble."
I think we are in a bubble, but bubbles have different shapes. I think the AI bubble will rhyme with the dot.com bubble. Most of the excess of the dot.com bubble might have been justified. If you look at the top market cap companies in the world, they include Amazon, Google, Paypal, eBay, Salesforce. All of these were started in the dot.com bubble. There are areas of excess today, but it would be dangerous to dismiss this as strictly excess, and thereāll probably be outsized returns within it.
- Brett Taylor
Hereās a snapshot of AI-first (?) companies revenue Vs valuation, you get the picture:
Combined valuation of >$17B
Combined revenue <$100M
2. The rise of the āAgentic Economyā is a potentially (very) big deal
Every major platform generational shift has prompted the rise of a new type of economy. Every time, they have given rise to huge, established tech players.
The same players that you have probably been reading about lately, about how they are scrambling to leapfrog to the next rising economy.
Here are a few examples youāve all experienced first-hand:
Public Cloud enabled the SaaS economy
The iPhone enabled the App economy
Social media enabled the Creator economy
LLMs gives rise to the Agentic economy
And hereās why it is potentially very exciting:
AI agents are changing traditional software ā Instead of the usual click-around menus, separate data, and pay-per-user model, AI agents do things differently.
AI agents reduce the need for human labor ā Companies spend a lot more on workers than on software (35x+). AI agents can help cover both areas, potentially saving on labor costs.
AI agents make services more efficient ā In industries where work is done by people and profit margins are low, AI agents can boost productivity and reduce costs.
3. There are 3 types of emerging AI Agents: Personal, Role-Based, and Company-Focused
āAutonomous agents are programs, powered by AI, that when given an objective are able to create tasks for themselves, complete tasks, create new tasks, reprioritize their task list, complete the new top task, and loop until their objective is reached.ā
āMatt Schlicht
Agents have the potential to be so useful that you might see them everywhere soon:
4. The way AI Agents are built challenges traditional software development practices
A new kind of software demands a new approach to development.
- Clay Bavor
As AI agents replace traditional rule-based software, they challenge established development practices by introducing unpredictability, high costs, and unique upgrade issues.
Adapting to this shift requires new strategies for reliability and future-proofing.
Digital Shift: Software development has evolved with structured best practices (SDLC) for reliability.
AI Agents Challenge SDLC: Unlike traditional software, AI agents use flexible, goal-based models, creating unpredictable results.
Input Differences: Structured input (forms) vs. natural language, leading to infinite interaction possibilities in agents.
Performance & Cost: Traditional software is fast and low-cost; agents rely on slower, more expensive LLMs.
Upgrade Instability: Traditional updates are smooth; LLM updates can disrupt agents, requiring re-training.
New Paradigm: Moving from deterministic, affordable software to adaptive, costly AI agents poses new reliability challenges.
5. Commercially, the AI market might play out like the cloud market
What learnings could we derive from how the cloud market played out, that might ārhymeā with what is currently going on in AI?
Well, 3 big categories emerged:
Infrastructure as a Service: Azure, Google cloud, AWSā¦
Tool makers: Datadog, Snowflakeā¦.
SaaS: Salesforce, Adobe, extremely long-tail of solutions.
The SaaS category has the most 2B-20B company outcomes, as represented in the stock market - and we might see something similar happen with this wave.
Much of the current media and VC focus is on data centers and foundation models.
The current āconsensusā is that this is the safest layer to be in. Group-think dictates that you know that whatever happens on top, those layers will ācollect taxesā. Butā¦
The idea that somehow the way the world wants to buy software will change becuase these foundation models are really smart doesnāt resonate.
āBrett Taylor
And if this wave does end up rhyming with the cloud marketā¦
6. Applications might be the most exciting area
Labor budgets are 35x software budgets, and AI agents eat into both.
A few takes on short-term impact:
šššŗš®š» + šŗš®š°šµš¶š»š² š³š¶šæšš: In the short term, a "human + machine" angle still wins. Industries with built-in failover mechanisms (like customer support) are ideal for initial AI deployment, allowing AI to handle basic tasks now and more complex ones over time.
š„š²š¶š»šš²š»šš¶š»š“ š£šæš¶š°š¶š»š“: AI agents challenge traditional SaaS models by moving beyond point-and-click interfaces, siloed data, and seat-based pricing.
šš š½š®š»š±š¶š»š“ šš°šæš¼šš š š®šæšøš²šš: AI agents are transforming industries by deploying across horizontal applications (cross-industry tools like legal AI for enterprises), vertical applications (industry-specific solutions like AI for law firms), and consumer services (full-stack services like legal assistance apps).
š¤šš²ššš¶š¼š»š š¼š» šŗš¼š®š: Rapid advancements in foundational AI models may commoditise a lot of early / short-term winners' use-cases. Raises questions on what sectors and verticals are more suitable to build product moats around. AIās accuracy in regulated industries is improving (these plays could have a significant moat).
7. An Agent Infrastructure layer is needed for the ecosystem to blossom
There are a few pre-requisites for this to take place:
Privacy is key ā Since agents handle a lot of personal data, they need strong privacy protections.
Efficiency and speed matter ā Running agents on devices directly (instead of remotely) keeps costs down, response times fast, and memory usage low.
Agents need to work together ā Agents should be able to communicate and work with each other smoothly, even across different platforms (like different calendar agents).
Memory for improvement ā Agents should remember past actions to improve over time and become more personalized, even anticipating needs.
Accuracy safeguards ā Built-in checks prevent agents from providing wrong or misleading information, helping maintain quality and trustworthiness.
Examples of scaffolding needed to build agents around large foundation models:
Auth: Anon secures agent identity, enabling potential automated actions like flight booking.
Security: Invariant Labs builds AI security for trusted, adaptive protection.
Frameworks: Langchain, LlamaIndex lead in agent app dev, with major F500 adoption.
RAG: LlamaIndex, Unstructured add real-time data for smarter agent responses.
Orchestration: Multi-agent systems delegate tasks for complex queries.
Runtimes: Modal, Browserbase offer scalable, low-latency environments.
Routing: Martian, DSPy optimize model costs and prompt efficiency.
Memory: MemGPT lets agents remember, personalizing interactions.
Evals: Weave (Weights & Biases) tests/improves agent performance.
No-Code: Brevian enables code-free agent deployment for businesses.
8. Agent adoption might come in 3 waves
Wave 1: Text-Based Agents ā Early adoption in text-driven roles (e.g., marketing, paralegal support, and customer service) where single-modal, language-based tasks are handled efficiently.
Wave 2: Multi-Modal Agents ā Expansion into complex, multi-sensory fields (e.g., architecture, gaming, and education) requiring more diverse skills and inputs.
Wave 3: Regulated Industry Agents ā Slower adoption in regulated sectors (e.g., healthcare and finance), where privacy is critical; success often hinges on building trust through advisory boards and compliance credibility.
9. We will see new Agent-development jobs
The web created jobs like UX designers and PMs.
AI agents will do the same.
AI doesnāt replace people, AI + human replaces people, raising the quality bar.
A few examples:
Agent Engineer: Building, deploying, and optimizing agents.
Agent Architect: Crafting effective and engaging agent conversations.
AI Workflow Designer: Designing workflows that use agents to automate tasks.
Ethics Officer: Ensuring agents behave ethically and comply with regulations.
Agent Integration Developer: Connecting agents to various tools to make them more useful.
10. How do you prepare yourself for this wave?
āEvery abundance creates a new scarcity. The task for builders is to figure out which resources AI makes abundant, which are rendered scarce, and where an edge can be formed. So far, for those who have found an edge, the business results are incredible.
Think about what resources does AI Agents make scarce:
Focused Attention: funny, considering Googleās paper that kicked this of.
Original Art & Writing: unique, unmistakably human creativity.
Trusted News & Facts: verifiable, AI-proof information sources.
Data Protection Skills: data privacy and security for AI-sensitive environments.
Customer Empathy: in-depth, nuanced, relationship-building skills.
Green Computing: energy-efficient, sustainable computing solutions.
AI Ethics and Governance: frameworks for transparent and ethical AI uses.
Complex Problem Solving: abstract reasoning and strategy beyond AIās capabilities (weāll see).
š Recent investments
At JME.vc, we're interested in this space and have already made several investments:
Hoola is developing a powerful AI agent shopping assistant for e-commerce.
Rauda is building a customer service agent capable of not only answering questions but fully resolving issues.
Genesy creates sales agents + data sourcing that can autonomously handle and enhance prospecting.
GPTAdvisor helps private bankers serve their clients more effectively and efficiently - at the intersection of agents and copilots.
A few more to be announced.
If you are building in this space, Iād love to learn about it - get in touch!
š AI Agents: 15 Must-Reads
š Pitch Magic AI: aka learning-by-doing
To my surprise, the Startup Pitch Deck GPT that I built on OpenAI now has over 25K+ conversations, >900 reviews and ranks n.1 for "startupsā in the GPT Store.
I wanted to experiment with this little toy and see if I should build a wrapper around it.
Iāve written about how to build a basic GPT and jailbreak it from OpenAIās wallgarden via their API before - and wanted to take it one step further.
But first, I want to pre-validate wether its worth building or not.
Thereās two possibilities:
Having the free GPT on the OpenAI store provides enough value, and might be monetizable when they decide to do rev-share or open a paywall feature.
People might be willing to pay for a fine-tuned, more advanced, pitch deck feedback co-pilot.
So, I started diving into emerging GenAI programming tooling.
And let me tell you, it is pretty mind-blowing how far you can get using ONLY natural language. Of course, you still need to fight bugs, and put a lot of hours into it - but Iām seeing tools every day that lower this barriers to entry at a phenomenal speed.
Hereās what Iāve gotten to:
Step 1: Write a PRD
Step 2: Whiteboard a landing page
Step 3: Develop a first landing page with natural language on v0.dev (above)
Step 4: Develop a plan with Claude.ai on what your architecture, user flows, and development stages will look like.
Step 5: Open Replit, start implementing your plan with the code provided by Claude (or any other tool you feel comfortable with). Hereās where the rubber meets the road and youāll have bugs and have to drive back, keep track of your conversations (because Claude has memory / conversation limits), and fight the good fight. Iāve gotten past building basic sign-up / login functionality, setting up Firebase and experimenting with forking a Github repo for the basic app-experience (very similar to ChatGPT + a library of deck assets potentially).
P.s. also found Bolt.new, which is sort of where this whole space is going towards (all-in-one solution for planing + building + deploying web app).
In my opinion, with steps 1-3 you should have enough to go out there and pre-validate wether something is worth building, or not (saving you countless headaches).
So, Iād be awesome to get your thoughts:
If interested in the free Startup Pitch Deck GPT go-ahead and enjoy (I appreciate a 5 star review if you like it) š!
š The reality of Product-Market Fit
People underestimate how much shipping it takes to hit product-market fit:
š EU-inc
Let's make Europe the best place for innovators.
Please sign this petition to create a pan-european legal entity: eu-inc.org
We have a once in a generation chance to improve Europe for startups.Ā
It's our job as industry to show how important it is.Ā
How you can help:
Sign the petition
Get your friends to sign
Share this in your whatsapp, slack and discord groups
Get this in front of press, newsletters, and influencers
šµ Iberian Deals
You love startups and want to enjoy a Spanish lifestyle?
Come join the Spanish startup ecosystem.
Hereās a list of recently funded startups:
Submer (data center cooling) raised $55.5M.
Koa Health (health) is preparing a ā¬40M round.
Paack (logistics) raised ā¬22M.
Tolerance Bio (biotech) raised $17.2M.
Matteco (cleantech) raised ā¬15M.
Be Levels (supplements) raised ā¬6M.
V2C (EV charging) raised ā¬5M.
Reveni (fintech) raised ā¬5M in debt and equity.
Depet (pet funeral services) closed a ā¬5M round.
GoTrendier (second-hand fashion) raised ā¬4.4M.
Ciudadela (property management) raised ā¬2.7M.
miResi (elderly care platform) raised ā¬2M.
Saigu Cosmetics (cosmetics) raised ā¬1.5M.
FABBRIC (fashion) raised ā¬1.25M.
Loud Intelligence (advertising) raised ā¬1M.
QUIXOTIC (energy) raised ā¬1.3M.
Zexel (payments) raised ā¬850k.
KomboAI (sales tech) closed a ā¬700k round.
Corium (fashion) raised ā¬500k.
Soluciones Industriales HAUX raised ā¬500k.
Anyformat (AI, unstructured data) raised ā¬520k.
Legit.Health (dermatology) closed a new round.
š« Meme of the month
If you enjoy Startup Riders, Iād really appreciate a share - see you next month! š¤
Such valuable content, Ivan! Thanks for that :)
Shared with my confounder. This will be amazing