top of page

Building Machine and Human Autonomy

Working at Valory: the values we use to build and grow our world-class team and technology

Valory’s original tagline was ‘architecting autonomy’. Since 2021, we have united a world-class remote team around a radical vision of human autonomy enabled by machine autonomy. Together, we have built industry-leading technology for machine autonomy, enabling communities to co-own AI by harnessing autonomous (AI) agents. This article outlines how we have built, and continue to grow, the original community behind this technology by committing and recommitting to autonomy. First, we define ‘autonomy’ as it applies to both the machines, specifically AI agents, and the team we’ve built, which can be seen as human agents. Then, we distill autonomy into the more specific, actionable values we use every day. Finally, we share how these values are applied at Valory in the hiring process and our daily work, helping us grow the autonomous team building uniquely autonomous tech.



Fig. 1 Valory’s logo 0.1, circa September 2021.


Defining ‘autonomy’, for humans and machines


A truly autonomous (AI) agent is a piece of software that can proactively ‘detect’ information, process and ‘orient’ to decide what to do with this information, then ‘deliver’ appropriate tasks (or entire outcomes) accordingly (DOD). Similarly, a truly autonomous team member can proactively detect information, process and ‘orient’ to decide what to do with this information, then deliver appropriate tasks (or entire outcomes) accordingly. This level of autonomy is one of the defining characteristics of Valory’s autonomous (AI) agents. In the same vein, we believe this level of autonomy is what has enabled our human team to harness the best of our collective skillset and deliver this autonomous tech. 

However, no matter whether it is a machine or human receiving such instruction, ‘detect’, ‘orient’, ‘deliver’ is quite a vague framework to apply to real work without further context. In fact, ‘DOD’ is a summary we surmised in retrospect to group the more specific and actionable ‘Values’ we have articulated through 2.5 years of working together. Our values were forged in several iterations out of the highs and lows, as we stopped to identify the underlying value at play that was making the difference. Thus, our values are a shorthand distilled to increase alignment, whether via praise or constructive criticism. Along with several existing frameworks from other battle-tested teams, these values are used daily in guiding each other on how to approach tasks, praising each other in our all-hands meeting, and in our 360 performance reviews of each other. 


Distilling ‘autonomy’ into values



Fig. 2 ‘Building Human & Machine Autonomy: a framework’ © Valory AG 2024


Valory was incorporated in 2021, as a fully remote company. As our 8-person team approached the first performance reviews, we wrote V0.1 a first, more generic draft of our values including ‘communication’, ‘proactivity’, and ‘ownership’. By the time we were 17+ team members in April 2022, we had built the foundation of our team and v1 of the Olas protocol and stack, therefore V0.2 included more idiosyncratic values like ‘clarity through writing’ and ‘push the autonomous edge’. 


  • ​​Push the autonomous edge We operate at and push forward the cutting edge of autonomous technologies. We evangelize the technological and social benefits of autonomy. We push for personal autonomy, enabling colleagues to take ownership of their work.

  • Get to simple, get to done Complex systems are built from simple systems. We prioritize getting simple, sufficient solutions in place as soon as possible. We have a bias to action – in a position of uncertainty, we pick a simple path to keep us moving forward.

  • Mission over ego We prioritize the goals of the organization over our individual wishes. Internal processes should serve our progress towards our mission, and not be a goal in themselves. Expertise trumps seniority.

After building Olas v1 we began to work with external partners, so V0.3 in April 2023 ‘done is when the customer is served’, assigning reviews to internal/external customers identified at the outset, which identified dependencies and closed feedback loops, instead of project/product owners. We also began incorporating LLMs into our day-to-day work across the company, so we added the value: ‘work smarter with AI’. 


  • Start with the end in mind In everything we do, we visualize the outcome at the outset. This helps us prioritize appropriately and estimate the value of what we set out to create.

  • Clarity through writing We value asynchronous communication. Writing keeps our thinking rigorous, our meetings productive and our motivations honest. We are what we write.

  • Actively update your thinking We hold strong opinions, but actively seek to disprove them with external evidence. If the hypothesis can be rejected or the opinion loses its strength, we swiftly correct course.

  • Done means the customer has been served Every task must have a customer: external or internal. If the customer has not accepted the delivery then we’re not done!

  • Work smarter with AI If the AI can do it we shouldn’t be doing it.

Then, v0.4 added ‘first principles, lasting impact’. In the context of the rise of LLMs, early adopters and soon the wider market scrambled to incorporate AI into products. We saw many products aim towards lower levels of machine autonomy than we prioritized, enabling correspondingly lower levels of human autonomy than our founding vision, where humans can outsource entire outcomes rather than simply tasks. Our own discussions about how to incorporate LLMs, therefore, distilled a founding principle of Valory; we strive to be led by first principles, rather than imitating the market. After all, had we imitated the market we never would have created the first open-source framework for decentralized autonomous agents in crypto in 2021. This thinking led to many of our innovations, from creating a platform to simplify our DevOps, to outsourcing lots of the marketing of Olas from our stretched team to a truly autonomous DAO via Contribute, an app we built using our autonomous agents that posts its own tweets autonomously.

  • First-Principles, Lasting Impact We build ground-up insights to deliver market-defensible, breakthrough technologies that stand the test of time. Deep creativity over shallow imitation.


In November 2023, we began to use ‘Detect, Orient, Deliver’* to group these now numerous values for our human team. Shortly later, as we ran the Olas Dev Academy, we realized that this was almost identical mental model to how we described the three stages of autonomous behavior: ‘perceive, process, act’. 



Fig. 3 An artistic impression of machine and human autonomy converging


Applying our values to work at Valory


We apply these values in all our work at Valory. Here we summarize how they are applied during the hiring process, and then after joining.


During the hiring process


Our commitment to machine and human autonomy in our products extends to our day-to-day ways of working, and hiring is no exception. Below is our hiring process, followed by what we look for throughout the process.


The current hiring process


Steps 2-5 are all remote calls, bar the project which is an async remote submission.

  1. CV & note - async review

  2. Phone Screen - 15m

  3. Role-based Interview - 1h 

  4. Project - maximum 3h

  5. Final interviews - 30m (with the hiring manager, plus each of the remaining 3 founders)

Assessing candidates, throughout the hiring process

  1. First and foremost, being able and excited to work in line with our values is a dealbreaker. We are not the right place for someone who seeks an organization with entrenched and rigid processes, ways of working, and structures. We are the right place for someone who is excited by the prospect of learning to embody, and perhaps even shape, our values as you build this world-class tech with its creators.

  2. Secondly, we need people who are excited to contribute to our vision. We reject candidates who do not demonstrate this via a personalized note or who haven’t done their research before phone screens. We are actually encouraged when people use ChatGPT in their applications (see value #9). However, if AI alone could do the job, we wouldn’t be hiring for the role. We can tell when you rely solely on it without adding your own perspective and will assume you may do the same in your work… Not a good sign.

  3. Thirdly, we hire people we believe can contribute individually to realizing our vision together. To test this, our hiring process includes technical interviews with prospective colleagues followed by a short take-home project, completed remotely. The projects allow us to mostly remove traditional criteria like university degrees and are discussed your final interviews with the management team, alongside culture fit questions. We still have short interviews with each of the founders to ensure both parties feel confident and excited that the other is a good fit. Though we have gradually introduced coordinators and, increasingly, product managers, we still expect everyone to contribute individually and have no team members dedicated solely to management.Once you reach founder interviews, let us know if you’d like to be connected with a team member to ask questions about what it’s like to work with us!


After joining

Of course, values alone can only guide the team member’s experience. Our values must also be reflected in our structures and processes, which currently include:

  • Token options in Olas - incentivizing team members’ performance long-term by enabling them to co-own the success they contribute to.

  • Biannual 360 performance reviews - where we reflect on ourselves, our colleagues and our reports/managers, evaluating their delivery and alignment with our values.

  • Remote socials, now via Virtual Reality - previously enabled by Gather and playing remote games, we believe those creating innovative products should get to play with (and perhaps even work) in them, too. 

  • Biannual in-person offsites - enabling our globally distributed team to have valuable face-to-face time, half dedicated to ‘work’ such as knowledge sharing by training each other or presenting on our work and team-building exercises, and half dedicated to touristic/cultural experiences together and having dinner together.

  • Hackathons and crypto community events - all team members experience hackathons either during our offsites or during crypto community events, enabling us to be exposed to our own and others’ tech, delivering together in a sandbox environment where we can take more risks.

  • Professional development budget - we also allow team members to request professional development opportunities they feel would boost their work, from courses to conference attendance and more.

  • Optimizing for delivery - unsurprisingly for a startup, we have found that the team’s stress is linked to our own and external perception of our traction, so honing in on our values and improving our delivery processes help create a healthy, autonomous community. For example, our weekly ‘Town Hall’ meeting with the entire company centers on sharing progress towards our high-level Objectives, broken into Key Performance Indicators with clear Key Results. Team members share their successes and learnings, linking this work to the customers they served and the Objectives they furthered. Management strategic insights and company-level feedback, and all team members can share praises for their colleagues, citing the values where they excelled.

Conclusion

We hope this article is a useful insight for all those curious about Valory values, from the broader principle of autonomy to the DOD framing and specific values.

In particular, we hope this article is useful to anyone interested in joining Valory, shedding light on the hiring process and beyond.

And finally, we hope this article is useful for operators and all those contemplating distilling their own values and applying them throughout their work. *Adapted from the OODA framework.

Comentarios


bottom of page