{"id":5586,"date":"2026-09-16T16:56:30","date_gmt":"2026-09-16T16:56:30","guid":{"rendered":"https:\/\/redmonk.com\/jgovernor\/?p=5586"},"modified":"2026-09-16T16:56:30","modified_gmt":"2026-09-16T16:56:30","slug":"agents-are-choosing-the-software-infrastructure-now-and-were-not-really-ready","status":"publish","type":"post","link":"https:\/\/redmonk.com\/jgovernor\/agents-are-choosing-the-software-infrastructure-now-and-were-not-really-ready\/","title":{"rendered":"Agents are choosing the software infrastructure now, and we’re not really ready"},"content":{"rendered":"

\"Set<\/a><\/p>\n

I have been waiting for my colleague Stephen to write something about agents as the new new kingmakers for a while now, and he finally posted something about that today –\u00a0 Agents: The new, New Kingmakers<\/a>. The post is definitely worth reading in full, not least for the history lesson about how we got here, but here is the core argument of the piece:<\/p>\n

There were and are limits, of course, to what the models can do and the mainstream software development market had to grapple for the first time with non-deterministic machines that would not only make mistakes, but lie about them and disobey instructions. But the direction of travel is clear: machines are getting better and better at writing code, and as a result they\u2019re writing a lot more of it. Just as importantly, they\u2019re not only writing the code, they are often choosing the programming language used, the libraries and frameworks leveraged \u2013 even the fonts in the user interface. In fact, they will often choose technology that they have been explicitly told not to use.<\/p>\n

All of which implies that the power dynamics of the industry, just as they tilted towards developers, are now shifting towards agents.<\/p><\/blockquote>\n

Stephen continues:<\/p>\n

Between the current model abilities and the wild asymmetries between different product categories, there isn\u2019t likely to be a single dominant approach. In some markets, AI-assisted human workflows will be appropriate; in others, it will be headless, agent-only infrastructure.<\/p>\n

What\u2019s critical is distinguishing between vendors building for a new agent-centric world, and those merely bolting AI-on because marketing told them to. AI-washing is not indicative of a product that will necessarily be correctly positioned moving forward.<\/p><\/blockquote>\n

This new dynamic is absolutely fascinating, not least because RedMonk has historically done a solid job of understanding developer choices and helping other people to understand them. Where is the signal now? It’s increasingly in the models and agents and decisions and suggestions they make. That certainly has implications for the work we as company do. But that’s because agent-first decision making has implications for the work all of us do – whether developers or platform engineers, security and compliance folks, marketing or sales people, boardroom execs, journalists, consulting firms. It’s not just the SDLC that changes – it’s the entire decision making and management infrastructure. It’s a lot easier to talk about guardrails than it is to enforce them. Developers have long ignored “corporate standards” – indeed this was a core thesis of Stephen’s book The New Kingmakers, which is about bottom up adoption. We would talk to CIOs and they had no idea what people were actually using to get the corporate work done.<\/p>\n

I remember the banking exec saying “MongoDB – it’s popping up everywhere in the organisation, but I don’t even know what license it is. I will tell you one thing – I am not going to run my bank on it”. I laughed to myself – “dude, you already are.” Developer choices are production choices, and today agent choices are production choices. And as Stephen points out in the piece, agents aren’t exactly renowned for doing as they’re told. They’re the new grassroots, and they’re going to create new winners and losers, just as developers made MongoDB a corporate standard without top down buy-in, let alone a mandate. And what happens as the agents make the choices, and humans are having less of an impact feeding the frontier models with carefully hand crafted blog posts, documentation, Stack Overflow responses and so on? So many questions.<\/p>\n

But when it comes to answers, there is no doubt that we need to consider agents as peers in IT-decision making. Stephen pointed to a few examples, and they are literally all around us. We recently signed up Fly<\/a> as a client, for example. Here is Martin Casado of Andreessen Horowitz talking about the company’s latest funding round.<\/p>\n

Every wave of infrastructure is defined by who its primary user is, and for the first time that user isn\u2019t necessarily a person. That changes what a computer has to look like,\u201d said Martin Casado, general partner at Andreessen Horowitz. \u201cFly has spent years building infrastructure for that shift and is giving agents real computers instead of disposable sandboxes. We believed in that path early, and we think Fly is the platform on which this gets built.<\/p><\/blockquote>\n

So…\u00a0 Computers for agents.<\/p>\n

I was about bring this positioning up with Stephen on Slack, when I noticed a new rhetorical linkedin post<\/a> by Render CEO Anurag Goel:<\/p>\n

We spent years hearing: \u201cWill Render scale?\u201d
\nOpenAI builds on Render. Base44 scales on Render.
\nMeanwhile, weekly paid service creation looks like this.
\nI think we can retire the question.<\/p><\/blockquote>\n

With a chart like this, that seems reasonable.<\/p>\n

\"\"<\/a>
post showing significant growth in Paid services created on Render.<\/figcaption><\/figure>\n

The first questions on the post asked: “what happened in January 2026”. To me the answer was obvious – capable autonomous agents happened. The period from November 2025 to February 2026 was when it became really obvious that we’d crossed a Rubicon. We could trust agents far more than ever before. The human in the loop became somewhat anachronistic. Organisations stopped worrying about manual code review, or at least realised that automated and AI supported verification and testing was going to be essential to shipping software in the new era. The acceptance of autonomy began with people playing around with pet projects, was quickly embraced by startups, and then moved into enterprise thinking far more quickly than I, for one, anticipated. I have spent all year talking to organisations that are much further along the autonomy spectrum than expected – including banks, telcos, retailers and government organisations.\u00a0 2025 may have been the year or agents but 2026 has definitely been the year of autonomous agents. Humans were the bottleneck in the software development lifecycle, but we’ve moving beyond that now. Autonomous agents do an awful lot of work – and that’s what the chart above shows. It certainly doesn’t hurt when you have OpenAI as a customer as Render does.<\/p>\n

Stephen in his post talks to the explosion of Sandbox platforms<\/p>\n

Consider the emerging Sandbox application category: from players like Cloudflare, the aforementioned Daytona, Docker, E2B, Modal and now Vercel, the entire market is an artifact of and would not exist without agents. Developer tooling built for humans was built on an assumption that they will operate within acceptable boundaries of behavior; developer tooling built for agents assumes the opposite. Sandboxes exist to give the exploding number of agents more autonomy, while not trusting them.<\/p><\/blockquote>\n

With agents making the choices technical and developer marketing disciplines are going to be fundamentally reworked. We’ll certainly see attempts, many successful, to game the system – GitHub stars for the agent era. SEO for software agents (GEO<\/a>). The zone will be flooded with vendor slop.<\/p>\n

All that said, I remain cautiously optimistic. Trust and authenticity and clarity perhaps have more value than ever – the need to understand and synthesise the choices and thoughts of smart engineers to get better work done. Models need good signal and inputs to have good outputs. Agents are the new new kingmakers, but humans are still very very good at systems thinking, architecture, and creative thinking. I plan to recommit to writing more to help stake my claim, and help others navigate, this new frontier.<\/p>\n

 <\/p>\n

disclosure: Cloudflare, Fly, GitHub, MongoDB, and Render are RedMonk clients.<\/p>\n

 <\/p>\n

 <\/p>\n","protected":false},"excerpt":{"rendered":"

I have been waiting for my colleague Stephen to write something about agents as the new new kingmakers for a while now, and he finally posted something about that today –\u00a0 Agents: The new, New Kingmakers. The post is definitely worth reading in full, not least for the history lesson about how we got here,<\/p>\n","protected":false},"author":5,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2},"jetpack_post_was_ever_published":false},"categories":[1],"tags":[],"class_list":["post-5586","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"acf":[],"jetpack_publicize_connections":[],"jetpack_sharing_enabled":true,"jetpack_shortlink":"https:\/\/wp.me\/p9wfjh-1s6","jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/redmonk.com\/jgovernor\/wp-json\/wp\/v2\/posts\/5586","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/redmonk.com\/jgovernor\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/redmonk.com\/jgovernor\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/redmonk.com\/jgovernor\/wp-json\/wp\/v2\/users\/5"}],"replies":[{"embeddable":true,"href":"https:\/\/redmonk.com\/jgovernor\/wp-json\/wp\/v2\/comments?post=5586"}],"version-history":[{"count":1,"href":"https:\/\/redmonk.com\/jgovernor\/wp-json\/wp\/v2\/posts\/5586\/revisions"}],"predecessor-version":[{"id":5589,"href":"https:\/\/redmonk.com\/jgovernor\/wp-json\/wp\/v2\/posts\/5586\/revisions\/5589"}],"wp:attachment":[{"href":"https:\/\/redmonk.com\/jgovernor\/wp-json\/wp\/v2\/media?parent=5586"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/redmonk.com\/jgovernor\/wp-json\/wp\/v2\/categories?post=5586"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/redmonk.com\/jgovernor\/wp-json\/wp\/v2\/tags?post=5586"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}