Lens Academy

Lens Academy aims to help 1 million people reduce x-risk from misaligned AI.

We get people excited about contributing, help them create a personal action plan, teach them the skills they need to succeed, and continue guiding them until they've achieved their goals.

We do so via scalable online courses and ongoing AI guidance.

There are many field-building organizations working on the common goal of preventing existential outcomes from advanced AI. Naturally, our activities overlap somewhat with those of organizations like BlueDot Impact and 80,000 Hours. We think that's great, and we are happy to learn from their success. At the same time, we bring a uniquely productized approach to capacity-building that we think is highly leveraged. We also fill gaps that other orgnizations aren't covering as much.

You'll read more about our vision and what makes us different in the "Theory of
Impact" section.

Team
When our founder, Luc, was a student, he founded SOLID, a renewable-energy nonprofit that's still thriving after nine years and has produced a commercial spinoff that raised over €100M. After years focused on climate change, he turned to the risks of advanced AI and founded Lens Academy in December 2025.

After help from some volunteers in our early months, Luc recruited a founding team of talented and mission-driven people, mostly former founders, who have started around June 2026. You can meet the rest of the team here.

Output
Your grant would act as bridge funding, enabling us to continue our operations until we secure a larger grant from CG or SFF. We're already part of SFF's main round and will hear back from them in September, and we will soon apply to Coefficient Giving.

We expect to rapidly grow our number of participants over the next months. Exact numbers and timelines are speculative, mainly due to the nature of product startups and exponential growth. Our expected range is 10k-200k course participants over the next year, though we’re internally aiming for a much more audacious 100k participants per month by Christmas 2026. Read more about how we're planning to grow here.

During the next months we will continue to scale up, improve our courses, and guide people to impact. We suggest not to see this grant as "X USD will enable N course participants" but as a bet on where we will be 6-12 months from now.

Lens Academy aims to help 1 million people reduce x-risk from misaligned AI. As a capacity-building organization, our impact is indirect: our impact is the extra impact our participants have because of us.

Our impact = number of participants x impact per participant.

As mentioned in the first question, we get people excited about contributing, help them create a personal action plan, teach them the skills they need to succeed, and continue guiding them until they've achieved their goals.

We can thus approximate our causal chain to impact by:

1. Reach many people
2. Get them excited about AI Safety
3. Help them figure out their path
4. Teach them what they need to succeed at their goals
5. Help them take actions with positive expected value

We can't predict exactly how our alumni will have impact. Instead, we'll teach them the strategic gear models to understand the landscape and empower them to find their own path.

There our two components of our vision that we want to zoom in on here:

Attract Talent at Scale
AI Safety needs more exceptional people, but only a small fraction of worldwide talent has even considered contributing. We think scaling up courses massively is a great way to reach more talent.

Much of our target audience would counterfactually not take action on AI Safety. Scale is important, because if more people learn about AI Safety, then more people will talk about it, causing even more people to want to learn about it. This creates a positive feedback loop, leading to more top talent entering AI Safety. Read more about why we focus on scalability.

We also don't have an application process, and accept 100% of enrollments. This better matches the early stage our prospects are in with respect to AI Safety, and supports our growth-feedback-loop. Read more about why we don't have an application process.

Besides exceptional talent, we also serve mainstream talent that gets rejected by other courses and fellowships that can't cheaply scale. Such talent is more likely to make part-time contributions than succeed at full-time career transitions. We're excited about the cummulative impact of small actions, but also feel uncertain about it. More thoughts here.

We enable our scalabililty through automation, volunteer facilitators, and AI guidance, achieving an estimated marginal cost of 30$ per participant.

Stick with them until they have impact
Our job isn't done when someone finishes a course. It’s hard for people to figure out how to get going in AI Safety. So we help people find their unique journey, provide further upskilling through advanced courses if needed, and continue to guide them as they navigate a landscape of courses, programs, and fellowships on their way to impact. It's our job to ensure they find a place where they take actions that reduce x-risk.

Here are three additional multipliers that increase our expected counterfactual impact:

We Focus on Producs
Products allow for outsized impact on the world because they uncouple outputs from inputs. This makes scaling both much cheaper and operationally easier, and is why venture capitalists love investing in product companies.

By contrast, we consider most fieldbuilding programs to operate more like services, where costs scale roughly linearly with outputs. While existing programs, bootcamps, and fellowships are very impactful and should absolutely be continued, we think scalable capacity-building products are boh highly leveraged and heavily neglected. We're excited to be building them.

We Have Good Feedback Loops
During our courses, we track engagement and learning outcomes via graded tests, time-tracking, and analyzing AI conversations and group meeting transcripts. This allows us to continually improve our program. After course completion, the continued guidance from our Lens Coach AI both increases the chance people actually take useful actions, and gives us insights into how we can best shape our program to help people reduce x-risk.

We Fill the Gaps, and Enhance the Rest
We identify gaps in AI Safety education and create courses to fill them. For instance, few programs were focused on AI Takeover risks, so we created a course around If Anyone Builds It, Everyone Dies, which we found highly effective at getting people to care about such risks.

Where existing courses work well, we partner with their creators to enhance and scale them. In such partnerships, our testing and AI Tutoring helps the courses be more effective, while our automated operations help them reach more people. We ran a demo partner course with AFFINE, and currently Effective Thesis is porting their curriculum over from Google Docs to our platform. We're having exploratory conversations with several other organizations.

Track Record
Luc founded Lens Academy in December 2025. With help from some volunteers, we started our first cohort of 20 students in January 2026. In the meantime, we built a fully custom course platform, allowing us full control over the learning experience. For instance, we built an AI tutor with scaffoldig for teaching AI Safety. It's fully integrated with our platform and has access to a database of thousands of curated AI Safety articles.

To be ultra-scalable, we're automating all our operations. We already have automated signups, group scheduling, Discord channel creation, Google Calendar invites, progress tracking, attendance tracking, participant grading, and a feature that enables participants to join another group for one meeting or permanently.

Based on our first experiences we decided to host a second small pilot cohort in March and April to trial another curriculum. During our pilot cohorts, we identified gaps in our ability to have scalable impact and decided to build Lens Coach, an AI that guides people to impact and also helps with course operations.

To start scaling up, Luc then recruited the rest of the founding team, who have started around June 2026.

We’ve also recruited 26 volunteers to facilitate the online group meetings of our participants, and we're currently running a beta cohort of 104 students spread across 9 groups. They're in week 3 of a 6-week course.

Our Lens Tutor and Coach AIs have received and responded to over 9,631 messages from 267 distinct learners. We've tracked 1,115 hours of studying on our platform.

Other things we've built:

• A facilitator dashboard for student progress tracking
• PromptLab: a tool for testing and iterating on AI system prompts
• A Zoom plugin that allows facilitators to monitor whether participants in breakout rooms are on the right track or need assistance
• Grading and Certificates: a participant scoring system integrated with LinkedIn certificates
• And much more.

The talent is out there. Let's get them excited, give them a path, skills, and guidance. And let's make it scale.

We're currently operating on an extremely tight budget, paying our team members so little that most are digging into their savings. This is a testament to the mission-drivenness of our team, but it is not sustainable.

This grant is bridge funding that would double our runway from 2 to 4 months. We're in SFF's main round (decision expected in September) and we're applying to Coefficient Giving as well.

This funding would enable:

• Runway doubled from 2 to 4 months, allowing us to focus on fundraising with CG instead of pursuing smaller but faster funders.
• Slightly raising the compensation of several team members to decrease their financial stress. These are small raises of <1k USD/person/month.
• Continued use of Opus 4.8 for our AI Systems, instead of switching to Sonnet and losing valuable training data. At a marginal cost of 30 USD/student/course, our program is very cost-effective, but we simply can't afford to continue using Opus much longer. Switching to a cheaper model would not only decrease performance now, but also prevent us from gathering important data that helps us refine the scaffolding and prompting of our AI systems.
• Occasionally contracting specialized talented individuals, such as for support with UX design or short-form video creation. We know people whose help we'd like and who are willing to work at very low rates to support our mission, but can currently not afford this.
• A small marketing campaign to get a baseline on paid customer acquisition cost, which will help us derisk our plans for growing.

Full breakdown in spreadsheet

At partial funding, we would reduce external contractor and marketing spend, potentially cap participant numbers, and if needed, attempt to raise more bridge funding from other donors.