← Dev & Engineering AI: Code Co-Pilots | Full Code Generators

M13’s RECOMMENDATIONS FOR

AI code co-pilots

Assisting engineers in writing code has been an key early use case for generative AI. Various tools are attacking the problem from different angles, with the goal of enhancing the output of highly expensive engineering resources. When done right, implementing code copilots can multiply the impact of teams large and small.

Updated 5/21/24 by ROB & ZACH


Maturity: ●●●●○ Partial Workflow Replacement

Startup Impact: ●●●●● Very High

  • Can supercharge a small core engineering team

  • Significantly improves average developer output

  • Huge bang for your buck, as engineering resources are expensive and hard to find

Areas of Improvement:

  • Limited understanding of context leads to suggestions that lack business logic or don't meet project requirements

  • Generated code outputs may not be performant

  • Humans are still necessary to test outputs and identify errors or security vulnerabilities

  • Legal concerns exist around intellectual property, alongside privacy concerns about the training data supporting these models

🔥 Our Top Picks:

GitHub Co-Pilot

Best for organizations with a wide range of technical teams looking for a universal tool.

  • GitHub Co-Pilot brings paired programming to development teams of all sizes, increasing efficiency with AI-powered suggestions. This functionality allows teams to reduce time spent on repetitive tasks, remembering syntax and building boilerplate code. As Co-Pilot's functionality has expanded, it also has allowed teams to onboard engineering talent faster, with conversational functionality on top of their own codebases, and to ship quality code faster by facilitating debugging and code reviews. With this combined functionality, Co-Pilot can be useful for a variety of team personas involved in the development process.

Codeium

Best for experienced teams looking to increase efficiency with a context-aware tool.

  • Codeium brings context-aware AI functionality to development teams, increasing the quality and relevancy of auto-complete suggestions. In addition to the quality of suggestions, Coedium has also added meaningful functionality to support teams of all sizes. Developers of all seniority stand to benefit from the Codeium Chat functionality, for both development workflows and as a knowledge resource. The relevance of suggestions combined with an intuitive user experience make Codeium a simple tool for engineering teams to adopt and see results quickly.

Fine

Best for resource-constrained teams looking to automate tedious or repetitive tasks with an agentic approach.

  • Fine's agent-based approach to development workflows allows engineering teams to call AI agents from systems like GitHub and Linear. Users are able to interact with an agent, asking for guidance in completing a task or asking the agent itself to create a PR with its own coded solution. Fine's agents learn from connected repositories, meaning its agents are currently best at making adjustments or addressing bugs with developer oversight.

AI full code generators

Unlike the code copilot use case, full code generation has the more ambitious goal of generating usable code completely autonomously based on prompting and other input, an extension of how no-code platforms generally work. The use cases are largely limited to front-end web applications and some basic mobile apps, but the usage is expanding rapidly. The issue with these platforms is the quality of code generated to date has been largely lackluster, though recent experimentation has yielded significantly better early results. An interesting space to watch!

Updated 5/21/24 by ROB & Zach


Maturity: ●●○○○ Highly Limited

Startup Impact: ●●●○○ Medium

  • Can potentially save huge sums of money for startups serving enterprise customers

  • Can potentially save lots of time in the deployment of traditionally engineering-heavy projects

Areas of Improvement:

  • The code generated is generally of subpar value

  • Engineers are unable to use the generated code and/or unable to easily edit it due to quality issues

  • Highly limited in use cases as startups are targeting specific over general solutions

🔥 Top Trending Tools:

Builder.io

Best for teams looking to experiment and iterate quickly on front-end experiences.

  • Builder.io is a versatile headless content management system (CMS) and development platform. It empowers engineering and marketing teams to create, manage, and optimize digital experiences without compromising on flexibility or scalability. Its embedded AI functionality allows users to use natural language in the creation of components, content, and mini applications, unlocking faster iteration cycles between marketing and engineering.

Goptimise

Best for early teams looking to ship quickly with light backend experience.

  • Goptimise allows teams to build APIs and databases in a low-code or no-code environment. The ability to ship a full backend in support of a product means teams can get feedback quickly and spend more time with design partners and customers. We are also excited about Goptimise's public roadmap, which previews a wider set of use cases.

Pico

Best for individuals building prototypes or simple web applications.

  • Pico uses text prompts that describe an application to build and deploy simple web apps. For a freelancer or individual looking to iterate quickly, this represents a faster option than other legacy tools.

Enjoying RocketGuides?

Have a burning topic you’d like us to explore? We're all ears! Share your thoughts and ideas by taking our quick, :30s survey. We’re eager to hear your feedback (good or bad!) to help us improve.

Rate M13's RocketGuides