Kilo Code Warns Developers: Token Costs and Vendor Deals Can Upend Your Stack

From punched cards to tokens

Rubiano opens with a story from his father, Eduardo, who studied industrial engineering at the Pontificia Universidad Javeriana de Bogotá in the 1970s. Students then ran code on IBM computers by feeding in ordered stacks of punched cards. The worst prank on campus, his father told him, was knocking someone's stack to the floor. Programmers could punch sequence numbers into columns 73 to 80 of each card so a sorting machine could restore the deck, but freshmen were not always that careful.

Rubiano draws the parallel directly: "tokens are our new set of punched data cards," he writes. This time, he argues, the threat is not a prankster but the cost of tokens and the changing terms of service of the companies developers rely on. Access to a tool, he says, is not the same as control over it.

Examples of dependence on a single vendor

The post points to Cursor as an illustration. According to Rubiano, Cursor's current Plus plan includes only Grok models, while third-party models on higher plans are billed separately by usage. He also writes that OpenAI proposed November 12 as the end of the agreement under which it supplies its models to Cursor. On October 29, Cursor reported a malicious extension posing as Claude Code in its marketplace, which was available for about 11 hours.

His point is that pricing, availability and model selection depend on others. A pricing change or provider disagreement will not break your code, he writes, but it can change how you build, whether you can afford it, and how hard it is to switch.

Currency swings and rising costs in Latin America

Rubiano argues that Latin American developers feel this most, because they often earn in local currency and pay token bills in U.S. dollars. He cites several figures:

• In Colombia, the dollar went from 4,478 pesos in November 2024 to 3,048 in August 2026, which he calls a 32% swing in under two years.
• In Brazil, the dollar jumped from 5.04 to 5.50 reais in a single day this past October.
• In Mexico, it climbed from 16.86 to 18.79 pesos between September 4 and October 30, 2026.

He adds that Gartner expects AI coding costs to pass the average developer's salary by 2028 as token consumption grows. Cost is not only a regional issue, he notes. In Stack Overflow's 2026 Developer Survey, 32.6% of the 1,033 developers who answered a question on cost and usage limits said they personally monitor usage, and 25.3% said they choose cheaper models when possible. Rubiano stresses that this is a global, self-selected sample, not a Latin American survey.

He also highlights the size of the regional talent pool. GitHub's Innovation Graph counted roughly 20.4 million developer accounts across 21 Latin American economies in Q2 2026, more than China's 11.8 million. Brazil had 10.2 million developers on GitHub as of September 2026, according to Octoverse 2026, and added 3.28 million this year, the third most of any country. These are GitHub accounts, he says, not a census of employed engineers.

Kilo's pitch: keep your options open

The second half of the post describes Kilo's products. Kilo Desktop by Anaconda offers a desktop workspace where users can choose from over 500 hosted models through the Kilo Gateway, use their own provider API keys, or run compatible local models. It also brings code, files, terminals, git and notebooks into one environment.

Kilo's individual platform is free to start with, with free model options, paid inference on a consumption basis, and purchasable Kilo credits. Teams on a ChatGPT plan can sign in with ChatGPT and use the usage that plan includes instead of paying per token through the API. Auto-Efficient sends each request to the lowest-cost model Kilo has measured as capable of that kind of task. Local inference avoids hosted token charges but demands proper hardware. "None of this makes the economics disappear," Rubiano writes, but it gives more ways to manage them.

He also cites OpenRouter's Chris Clark, who says open-weight models now carry about 64% of the tokens but only 28% of the spending. Routing simple decisions to smaller, cheaper models and saving frontier models for harder work is one tactic he mentions.

What it means for developers

The practical advice is to test rather than commit. Rubiano suggests downloading Kilo Desktop, running the same real coding task with two models, and comparing results and cost. Developers still need to check output, understand trade-offs and review how a provider handles their data.

There is also a commercial offer. Companies using and paying for Cursor can claim Kilo Enterprise free for the rest of that contract, up to a year, if they do so by November 30. As the post comes from Kilo itself, its claims about its own products are the company's view. The underlying argument is that model choice is now a budget decision as much as a technical one, and that developers who keep alternatives within reach are better placed when a vendor changes its terms.