On Friday, Frontier announced Catapults for EVE Frontier: one-way gates that fling your ship across gaps the Cycle 6 stargate network cannot reach. They are expected in game on Tuesday. By the weekend, EF-Map already had a working planner for them. This is the story of Yeet Planner, EF-Map's Catapult infrastructure planner, how it was modelled from announced rules before the patch landed, and why the fast first version was the least important part of the work.
A new travel problem
Cycle 6 removed jump drives from normal ship travel. If a system is not connected to the stargate network, you simply cannot get there, and roughly 20,000 of the 24,000 or so systems in EVE Frontier have no stargates at all. The announced fix is a pair of temporary buildable tools: the Mini Catapult with a 65 light-year range and the Heavy Catapult with a 365 light-year range. Unlike Mini and Heavy Gates, Catapults are one-way. You yeet your ship at a destination and there is no return ticket unless you plan one.
That changes the question players need answered. It is no longer "what route do I fly?" but "what do I need to build, and where, to make this trip possible at all?" A normal route planner is the wrong shape for that question. What you want is an infrastructure plan: the sequence of systems, which hops use existing stargates, and where you would place Mini or Heavy Catapults to bridge the gaps.
Modelling Catapults before they land
The announcement came with an early excerpt of patch notes: ranges, build costs, and one intriguing limitation about interference from nearby stars. Catapults are not live yet, and none of this data exists in the game client we can extract from. But the rules as written were specific enough to model, and EF-Map already had everything else: precise star positions for the whole universe, the stargate network, a routing engine, a Route Systems window for displaying hop-by-hop output, and share-card tooling from earlier features.
So instead of waiting for Tuesday, we built the planner against the announced rules and labelled it accordingly. When the patch lands, we extract the real data, verify the mechanics, and adjust. Players get several days of planning head start, which matters when the first thing everyone will do on patch day is race to claim useful staging systems.
Based on announced rules
Everything in Yeet Planner is modelled from the announced Catapult rules: 65 LY Mini range, 365 LY Heavy range, and the interference limitation as we interpret it. The rules may change when the patch ships, and we will verify against extracted game data once Catapults are live. Treat plans as strong drafts, not gospel.
The midpoint-sphere rule
The most interesting part of the announcement is the interference limitation. Without two gates resonating with each other, Catapults are weak to interference from other stars between you and the destination: stars within range can only be selected as the destination when there are no other stars closer to your current location.
Our working interpretation is a classic piece of computational geometry. For origin star A and destination star B: draw the straight line between them, take its midpoint, and draw a sphere centred on that midpoint with a radius of half the A-to-B distance. If any other star sits inside that sphere, the direct Catapult hop is blocked. If the sphere is empty and B is within range, the hop is valid. Graph theory fans will recognise this as a Gabriel graph edge test. Everyone else can picture it as a keep-out bubble sitting between you and your target.
The neat consequence is that a blocked direct hop is rarely a dead end. If a star blocks your line, you can usually yeet to that star first and continue from there, so the planner searches for chains of valid hops rather than giving up. Under the hood this is the same family of pathfinding we covered in A* vs Dijkstra, running over Catapult candidate edges generated on demand from star positions, with the interference test applied to each one.
From first pass to usable tool
The first working version of Yeet Planner came together quickly. The app already had the hard parts, and LLM-assisted development is very good at assembling a new feature from existing rails: a planner module, a web worker, a new Routing tab, tests for the geometry. That speed is real, but it is also the misleading part of the story, the same lesson we keep re-learning on features like RootFit. The first pass is a prototype wearing a feature's clothes.
What made Yeet Planner actually good was the loop after it: smoke testing, honest looks at screenshots, and player feedback once it was live. In rough order:
The polish that mattered
- Making it look native. The first UI was functional but visibly bolted on. It was rebuilt to match the Point to Point tab's exact layout rhythm, spacing, and advanced-options pattern.
- Removing clutter. Explanatory paragraphs moved to the Help panel. A Plan Mode dropdown turned out to never change the chosen route, so it was removed entirely.
- Route details in the right place. Hop-by-hop output moved out of the form and into the Route Systems window, where Point to Point routes already live.
- Optimisation wording players asked for. "Lowest build cost" was dropped because EF-Map should not pretend to know your economics. A new Prefer Catapults mode came straight from feedback: some players would rather build temporary Catapults than wake dormant stargates, so the planner now supports both philosophies, plus a Fewest Catapult Hops mode.
- Failure messages that help. Early failures quoted the direct origin-to-destination distance, which was true but useless. The planner now diagnoses why a plan failed and names the specific gap, for example that one hop needs an existing stargate because it spans 943 light-years.
- Sharing, added after we noticed the gap. Plans can now be shared as links, copied as summaries, exported as cards, and copied as in-game notes.
None of that list is glamorous, and all of it is the actual product. Fast prototyping earns you the right to start polishing early. It does not replace the polishing.
Why the planner avoids showing costs
The announced build costs are known, and Yeet Planner deliberately does not display material totals. Two reasons. First, EVE Frontier is in alpha and costs can change, and a tool that confidently displays stale numbers is worse than one that shows none. Second, the real cost question is a judgement call the player should own: waking a dormant stargate may cost more than building a Catapult, but a stargate is permanent two-way infrastructure and a Catapult is a temporary one-way tool. EF-Map shows you the shape of the plan, the hop counts by type, and lets you do the value maths yourself.
Sharing infrastructure plans
An infrastructure plan is a group activity. Someone scouts it, several people build it, a tribe uses it. So plans are shareable the same way Point to Point routes are: a short link that reopens the exact plan in EF-Map, a plain-text summary for Discord, a share card image, and an in-game note with clickable system links plus markers for what to build or wake at each hop. Stargate hops carry their community-reported status, so a hop marked wake required tells your fitting-out crew exactly what they are flying into.
What still needs verifying after Tuesday
Once Catapults are live and the data is extractable, the checklist is: confirm the Mini and Heavy ranges, confirm the build costs, and most importantly test whether the in-game interference rule matches our midpoint-sphere interpretation. Fuel, reuse, and upkeep mechanics are not modelled at all yet because nothing about them has been announced. If the shipped rules differ, the planner logic is isolated in one module precisely so it can be corrected quickly.
What this says about building EF-Map with LLMs
This feature is a clean case study in where LLM-assisted development actually earns its keep. It did not invent anything. It moved fast through well-understood territory because the architecture already existed: star positions from the system database work, routing patterns from years of pathfinding features, display and sharing rails from previous tools. The same audit-and-iterate workflow we use everywhere did the rest: prototype quickly, smoke test honestly, ship, listen, fix.
The result is that when a game announces a mechanic on Friday, players can plan for it over the weekend. If you fly in EVE Frontier, open EF-Map, head to Routing, and pick the Yeet Planner tab. There is also a Catapult planner overview page if you want the summary before diving in. Plan your yeets before Tuesday, and tell us what breaks.