Charting the Course: Why AI Needs a Steering Wheel
AI agents promise incredible efficiency, but the way they’re built today is fundamentally flawed.
Right now, agentic tools treat planning as a black-box process. You hand them a goal, and they “teleport” you straight to a final destination, without transparency into the small decisions they’re making along the way.
This teleportation might seem appealing at first. But every hidden decision is a potential wrong turn, nudging you silently off-course. By the time you realize something’s wrong, you’ve arrived at a destination you never intended to reach.
We think this black-box approach is broken.
Agents shouldn’t hide planning, they should surface it.
This means making the proposed strategy and key steps visible, not necessarily exposing every internal thought, but revealing the intended path so humans can engage with it. The goal is to enable humans and AI to shape the plan together, upfront, before committing to execution.
Humans should be able to inspect, refine, and steer plans collaboratively, guiding AI rather than being blindly guided by it.
But this desire for collaborative steering highlights a related challenge we’ve long faced even without AI.
The Human Side of the Planning Challenge
Staring at a big goal, the hardest part is often mapping out the first steps. Yet, most ‘planning’ tools today only help manage tasks after you’ve done that difficult upfront thinking.
Task trackers like Jira or Linear help organize work, but they don’t tackle the core planning challenge: figuring out the right tasks to begin with. This requires deep domain insight to define initial steps and, critically, structure them in a way that surfaces risks and allows you to learn quickly. They essentially presume the hard planning work is already done.
Breaking down work is one part of the challenge, but good planning also requires iteration. You rarely know the full path from the start. Perfect plans don’t exist. What works is starting small, learning quickly, adapting on the fly, and refining direction as you go.
Tools like ChatGPT can brainstorm tasks, but the result is typically a list embedded in a chat history, not a living plan. Trying to evolve this list through re-prompting is often cumbersome, risking context loss or inconsistent changes. It doesn’t function as a dynamic source of work that adapts fluidly as your project progresses, making genuine iteration difficult. On top of that, knowing how to prompt effectively to guide this evolution is often unclear.
This difficulty in human-led planning, compounded by inadequate tooling, mirrors the core issues we see with AI agents: a fundamental lack of support for transparent, context-rich, and adaptable plan creation.
Bridging the Human-Agent Gap
This vision is why we are building Fission.
We believe planning should be transparent, iterative, and collaborative for both humans and AI. But planning alone isn’t enough, it’s ultimately about getting things done.
Our vision for Fission is that it won’t teleport you. Instead, our goal is for Fission to:
- Visually map out AI-generated plans as dynamic, structured task graphs you can interact with.
- Enable collaborative refinement, letting you guide AI suggestions to iteratively adapt plans in real-time.
- Establish clear checkpoints with persistent, readily accessible context, ensuring you always understand the ‘why’ and remain in control.
- Transform ambitious goals into actionable, adaptable tasks through guided decomposition.
- Build a persistent, structured knowledge base for your project, ensuring context flows seamlessly between humans and AI agents without manual refresh.
Fission won’t be just a planning canvas; we are designing it as a launchpad for action.
The structured tasks and context will become the direct input for execution, whether you assign them to team members, tackle them yourself, or hand tasks off to AI agents to carry them out, closing the loop between intent and outcome.
Validating Our Approach
We’re not just theorizing about these challenges. We see people grappling with them every day as they try to make AI useful for complex work. The frustrations and the ingenious but often tedious workarounds shared within AI communities online paint a clear picture:
Users describe hitting roadblocks fast when AI operates like a black box, warning how “small decision nodes wrong… multiply that by ten and you’ve already landed at the wrong outcome.”
They wrestle with context. AI often overlooks critical details or needs exhaustive instructions. Some developers resorted to “markdown files that serve as your AI’s memory” just to keep projects coherent.
So what happens? People manually build the guardrails AI lacks. They invest hours upfront because, as many have learned, “If you don’t plan ahead and constrain it, the output will suck.”
They meticulously “develop a research plan first, then execute that plan step by step.” They try to force interaction, telling AI to “stop and wait for input after every step,” and want tools that let them “influence and correct these nodes on the go.”
That’s why we are building Fission. Not to patch the pain, but to solve it at the root. Our aim is to give people the integrated planning and steering layer they’ve been cobbling together on their own.
The Future of Planning and Doing
We’re entering a world where humans and AI agents need to collaborate seamlessly. Transparent planning and execution are critical, not just for achieving your goals efficiently, but also for ensuring alignment between your intention and AI execution.
Fission is being designed to be the interface that connects your intent directly to execution. Your structured, adaptable source of work, built for both humans and AI.
Let’s Build the Future Together
If this resonates with you, whether you’re tired of AI teleporting you or frustrated with the gaps in today’s planning tools, join us.
Together, let’s create the planning and execution layer that bridges the gap between human intuition and agentic action.
We’re actively looking for people interested in this problem space to share their experiences and help shape Fission’s direction!
- Share your perspective: If you’re tackling these planning challenges, we’d love to learn from your experience. Reach out to the team at [email protected].
- Stay updated: Join the waitlist below to hear about our progress.
- Follow along: See our journey unfold on LinkedIn.