1. Introduction to Kilo Code and Setup [0]

thumbnail

1.1. What Kilo Code Is [0]

Kilo Code is introduced as an AI agent, not a code editor itself, which the presenter plans to use within VS Code or Cursor.

  1. Tool Identification [0]
    1. The presenter is working with Kilo Code, which is described as an AI agent [0].
    2. This agent will be used in conjunction with Cursor or VS Code [1].

1.2. Account Creation and Initial State [4]

The presenter successfully created an account and noted the initial free credit allocation before installing the necessary extension.

  1. Account Creation Process [4]
    1. The presenter returned to the main website after creating an account in the top right corner [5].
    2. Login can be completed using a Gmail account [7].
    3. Upon logging in, the interface appears, and the user is granted five credits [8].
    4. The extension has not yet been installed at this point [9].

1.3. Extension Installation and Free Trial Activation [10]

The next step involved installing the Kilo agent extension and choosing the free trial option to begin testing.

  1. Installation Steps [10]
    1. The user clicked to install the extension and open the URI [11].
    2. The extension is identified as Kilo agent (Key loop put for free) [13].
  2. Choosing the Free Option [14]
    1. The user noted the option to "Use your own API" but chose to "try for free" [14].

2. Testing Kilo Code Performance and Cost [17]

thumbnail

2.1. Preparing for the First Task [17]

Before submitting the first task, the presenter closed Copilot and configured the AI model settings.

  1. Configuration Steps [17]
    1. The presenter closed Copilot to proceed [17].
    2. A task was prepared to see how the agent performs [17].
    3. Several models are available for selection [18].
    4. The presenter kept the default Kilo Code model selected [19].
    5. The API key was noted as already integrated, using the Anthropic Claude 3 model format [19].
    6. The settings were saved, and the system confirmed completion [20].

2.2. Analyzing Credit Consumption [25]

The initial five credits were confirmed, and the presenter calculated the approximate number of prompts possible based on the cost of the first query.

  1. Initial Credit Check [25]
    1. The total starting credit was confirmed as five credits [25].
  2. Cost Per Prompt Calculation [26]
    1. The cost per prompt was estimated to be around 0.05 or 0.06 credits [27].
  3. Estimated Total Prompts [29]
    1. Based on this cost, the user calculated they could have around 100 prompts using one Gmail account [29].
    2. The presenter was unsure if this limit was monthly [31].

2.3. Executing the First Design Task [32]

The first test involved designing a data table, which Kilo Code completed with impressive speed.

  1. The Prompt [32]
    1. The task was to "Design a data table with two data showing" [32].
  2. Performance Observation [36]
    1. The process was described as "too fast" and "seriously too fast" [36].
  3. Output Review [39]
    1. The result displayed a table titled "student performance data" [39].
    2. The columns included student name, GPA, science mark, and total marks [39].
    3. The presenter expressed confusion about how the total marks were calculated relative to the GPA of 3.9 [40].

3. Advanced Testing and Tool Evaluation [41]

thumbnail

3.1. Attempting an Advanced Problem [41]

The presenter moved to a more complex task involving debugging or solving an advanced programming issue.

  1. New Prompt Preparation [41]
    1. The user decided to ask for something "different" and "advance" [41].
    2. The specific advanced task involved a scenario where a Chemistry list was stuck in a loop [43].
  2. Agent Response to Complexity [45]
    1. The AI immediately suggested that the problem might be with its current strategy [45].
    2. The presenter waited to see how the agent would handle this new prompt [46].

3.2. Kilo Code's Utility for Problem Solving [47]

The presenter explained that Kilo Code is useful for debugging because it can analyze multiple files simultaneously within the IDE environment.

  1. Problem Solving Capability [48]
    1. This AI can be used directly for problem solving and resolving errors [48].
    2. Because it functions as an IDE agent, it can take all relevant files together [49].
    3. It then performs an operation to solve the problem from inside the environment [49].
  2. Result and Impression [51]
    1. The output included fields like TGP, physics, chemistry, biology, total mark, and CTP [54].
    2. The presenter felt the output was "too much" but acknowledged the tool's functionality [51].
    3. Overall, the presenter stated they "don't know" if it is the best, but it is usable [55].

3.3. Comparison with Competitors [55]

The presenter offered a personal recommendation, favoring other tools like Cursor and Visual.dev over Kilo Code for specific use cases.

  1. Personal Preference [55]
    1. The presenter stated that Cursor is best based on personal choice [56].
    2. Cursor is expected to produce "something unexpected" for the same prompt [56].
  2. Visual.dev Mention [57]
    1. Visual.dev was highlighted as another master tool, described as being on the "next level" for design tasks [57].

3.4. Final Credit Review and Conclusion [59]

After several prompts, the remaining credits were checked, confirming the low cost of the initial design task.

  1. Credit Usage Summary [59]
    1. After many prompts, 4.39 credits remained [59].
    2. The initial design task was completed using only 0.66 credit [60].
  2. Closing Remarks [58]
    1. The presenter concluded the overview of Kilo Code, hoping the audience liked it [58].