Challenging, chaotic, interesting, surprisingly similar to other great companies
Pros
Valve offers extremely generous benefits and perqs, and affords employees high levels of trust to do whatever they need to be productive. It is a privilege to work with the folks at Valve because nearly all are exceptionally accomplished, competent and eager to build something great. The environment really encourages employees to be positive and to focus on work that will directly impact the customer. Productivity is rewarded in part by peer review which makes employees accountable to their team. Changing teams/projects is usually easy, and is usually each employee's own decision. Employee autonomy is inherent in Valve's process.
Cons
Many of the ways in which Valve seeks to differentiate from other companies are not actually so valid. While it's true that Valve has no official job titles or promotions, compensation varies greatly among employees and many teams have an obvious pecking order. There is no formal management structure, but it's clear that some people have substantially more control over project direction and the work of others. Even though productivity is said to be the only metric that matters, people who are already connected or are accomplished social engineers will do just fine. Denying that all of these social forces are at work makes the problem intractable and difficult to even discuss. For a company that makes so much money, Valve is surprisingly risk-averse. New projects, internal tools, dev infrastructure, and anything that doesn't contribute to a current product are met with disdain. Because teams are intended to be self-forming, it's rare that enough people will want to assume risk to all collectively embark on a new project. It's too safe and too profitable to just contribute to something that's already successful. Even though failure is supposed to be tolerated and even encouraged so that employees will try new ideas and experiments, there is little evidence of this. After a few rounds of bonuses, folks learn quickly what is rewarded, and what is not. Valve's success has made folks arrogant, and this contributes to the problem of how new ideas are considered and discussed. Dogmatic thinking is actually common because people can always point to a great success in the past and use this to justify why everything should continue as it is. Some folks at Valve do not want the company to grow. Valve already has an incredibly strong profit/employee ratio. Why dilute it? This line of thinking crops up in project discussions as well, and causes many ideas to be dismissed because they seem too niche/unprofitable (at the time).