Months Spent on Bots Yield No Results | LLMs Best for Text and Code
Edited By
Fatima Rahman
A significant trend is emerging among developers as frustrations mount over the limitations of current AI capabilities. Many have spent considerable time attempting to build autonomous bots, only to find that they lack the effectiveness promised by advocates. This gap between expectation and reality is shaping a new perspective on AI’s role in the tech landscape.
The Testing Reality
In a recent test of Figma's AI functionalities, one user noted that components unpredictably "fall apart" and tokens get incorrectly utilized despite strict guidelines. Their insights illustrate a larger issue: while large language models (LLMs) excel in writing and coding tasks, they struggle with graphics and design interventions, emphasizing their current limitations.
"The hype around 'make a billion per second with AI bots' is noise from people who don't actually do this work."
This user’s experience reveals a growing consensus that the notion of LLMs functioning as autonomous agents is far from reality.
Diverging Opinions on Autonomy
In discussions across various forums, differing opinions surfaced:
One commenter stated, "True autonomy is nice for demos but terrible for a product." They highlighted the steep development time needed to address edge cases in autonomous systems.
Another added, "Tight constraints and explicit scope are the pattern—autonomy doesn't ship."
This feedback echoes a reluctance among many developers to embrace full autonomy for AI tools, recognizing them instead as valuable collaborators rather than replacements.
The Effective Workflow
The prevailing sentiment suggests a shift towards a more pragmatic approach. Developers are finding more success by setting clear specifications before moving to code, especially when engaging with tools like Claude. Users emphasize that “clear input” leads to “reliable output,” which is crucial in building trustworthy applications.
"What actually works: spec first, then code."
This change in strategy points to a promise of improved outcomes without the complications that come with overly ambitious AI autonomy.
Key Insights
🔍 Many have abandoned the autonomous AI dream, finding more reliability in collaboration.
⏳ True autonomy in AI is still several years away, and not ideal for production.
💬 "Claw is basically just agents with a cron" reflects a more simplified approach to automation.
As 2026 unfolds, it appears the conversation around AI continues to evolve—and users are clearly defining it through their experiences and challenges. The trend indicates a pivoting back to basics, focusing first on this:
Specifying needs
Harnessing LLMs for their strengths
Building user-ready outputs instead of chasing elusive autonomy.
For those developing AI tools, this insight could define how future projects are shaped—reinforcing the idea that current technology is best kept as a supportive resource.
Shifting Dynamics in AI Development
There's a strong chance that developers will continue to prioritize collaboration over autonomy in AI tools, aiming for clearer outputs. Experts estimate about 70% of teams will lean towards defined specs and task-oriented approaches in their workflows. As the tech industry recalibrates its expectations, we may see an increase in tools that integrate LLM capabilities while maintaining human guidance in the creative process. This shift could reshape product development and lead to innovations that focus on enhancing human creativity rather than striving for full autonomy.
Uncharted Waters of Innovation
Reflecting on the early days of the internet, many companies chased the notion of providing everything online without fully understanding the underlying infrastructure needed for success. Remember how early attempts at e-commerce resembled less than smooth experiences due to a lack of reliable payment systems? Similarly, today’s tech enthusiasts are learning to navigate the complexities of AI development, recognizing that dependable foundations are essential to building tools that truly serve their purpose, rather than getting lost in the allure of autonomy.