THE DRIVING FLY — PRESS KIT
Updated 11 September 2026
https://thedrivingfly.com/press
Press contact: Mark Unthank, mark@reallynice.company

PROJECT SUMMARY
A neural model built from a fruit fly’s measured brain connections learns to operate a simulated body at the wheel of a car.

BACKGROUND
The Driving Fly is an independent experiment by Mark Unthank at Really Nice. It uses measured fruit-fly neural connectivity as the structure for a trained model, connects that model to a simulated fly body, and uses the body’s measured controls to drive a Mini in CARLA. The project documents steering, indicators, horn use, parking attempts and three-point turns through films and open results. Current tasks use structured inputs or scripted requests, with engineered motor assistance; they do not demonstrate driving from vision. Sponsorship placements on the virtual Mini help fund the experiment. The code project is also known as Flyhard.

ABOUT MARK
He quit a $500k/year AI job to build Really Nice.
Mark Unthank is the founder of Really Nice and the creator of The Driving Fly.

Before starting Really Nice, he worked as a founding design engineer at an AI startup.

The Driving Fly is the kind of project he wants to make in public: a strange, ambitious experiment with the evidence left in. The films show the wins, the failures and the next problem to solve.
Visit Really Nice: https://reallynice.company

A NOTE FROM MARK UNTHANK
“The next step is to train it up and put it through a driving test in the simulator.”

SOCIAL PROOF
4M combined views
Launch film: The fly is learning to drive
https://x.com/alright_mark/status/2097800377080889409

Indicator film: Learning to use the indicators
https://x.com/alright_mark/status/2097963149794197910

QUICK FACTS
Project: The Driving Fly / Flyhard
Created by: Mark Unthank · Really Nice
Model: 165,122 traced MaleCNS neurons
Simulation: CARLA + MuJoCo / NeuroMechFly
Status: Independent experimental prototype
Funding: Sponsor placements on the virtual Mini

FILMS & CONTEXT
A little change of direction.
Three-point turn | 22.8 seconds | Boccherini | Normal simulation speed · 60 fps
https://thedrivingfly.com/media#three-point-turn
Download: https://thedrivingfly.com/media/three-point-turn.mp4

The full successful validation take plays once, without skipping ahead: 20.3 seconds of driving, followed by a short hold on the finish. The fly chooses steering, speed and gear from structured road and goal geometry. Fixed motor helpers move its legs; the measured wheel, pedals and selector operate CARLA.

This take finishes 25.4 cm from its target and 0.755° from the intended heading, with no collision. In eight separate held-out tests, six met the full stopping target. All eight completed the forward–reverse–forward sequence without contact or crossing the marked road boundary; two stopped just outside the position tolerance.

These are small, scoped tests. There is no camera-based road understanding, and the model is an engineering interpretation of measured fly connectivity. Boccherini’s Minuet and Trio accompanies the complete attempt.

Small fly. Big feelings.
Horn experiment | 20 seconds | Recorded car horn | Edited at normal simulation speed · 60 fps
https://thedrivingfly.com/media#road-rage
Download: https://thedrivingfly.com/media/road-rage.mp4

The fly passes one car, collides with the next, and keeps holding the horn as a third joins the pile-up. Interior cuts show the recorded foreleg movement operating the button. The real horn recording is timed to measured button presses.

This is a failed take, kept because the failure is part of the experiment. The source travelled 60.04 metres and produced two horn onsets, missing the test’s required 90 metres and three onsets. Its longest physical press lasted 16 seconds. The impacts and continuing hold remain in the edit; four seconds of waiting are removed.

Traffic, route, speed and erratic steering requests are directed. The model supplies the measured wheel and horn actuation. The cabin fly is composited from the same recorded body motion; this is not a demonstration of autonomous driving.

Green means go. Apparently.
Horn experiment | 37.5 seconds | Recorded car horn | Edited at normal simulation speed · 60 fps
https://thedrivingfly.com/media#horn-etiquette
Download: https://thedrivingfly.com/media/horn-etiquette.mp4

Cars arrive at the lights, wait, and pull away. The fly honks 0.117 seconds after the light turns green behind a waiting car. It drives through an already-green light quietly, then meets cars pulling out at a T-junction and a crossroads. Interior shots show it holding the horn.

The empty-road scene includes a brief, 83-millisecond false chirp. We kept it. None of these five source takes recorded an ego-vehicle collision, but a separate ten-case structured diagnostic passed only three of four positive horn cases and all six quiet cases. That small diagnostic does not guarantee quiet driving in CARLA.

The learned controller operates the fly’s forelegs and physical horn button using structured inputs. Approaches, traffic, route and speed remain directed. The cabin body is a synchronized composite, and the recorded horn sound follows measured button contact.

Fifty ways to miss a parking space.
Parking experiment | 30 seconds | Mozart | Accelerated montage · 60 fps
https://thedrivingfly.com/media#parallel-parking
Download: https://thedrivingfly.com/media/parallel-parking.mp4

The edit builds to 32 views by 6.4 seconds, then jumps between individual attempts and the grid. All 50 trials appear as excerpts. The grid footage runs at 6× speed; the fly alongside is a separately looped replay, labelled in the film, and is not synchronized to each car.

The result is 0/50 successful parks, with collision flags in ten trials. Resetting the learned core also produces 0/50, with twelve flagged trials. Flags include virtual-curb and vehicle-bound overlap as well as native collisions; they are not a count of individual crashes.

The policy chooses steering, speed and gear from structured relative geometry, with fixed motor helpers moving the body. The measured controls drive CARLA. Lower final-position error than the reset comparison is useful evidence, but it does not establish successful parking. The original sponsor snapshot is preserved.

A small indication of progress.
Indicator experiment | 25 seconds | Silent | Slow motion · 60 fps
https://thedrivingfly.com/media#indicator-closeup
Download: https://thedrivingfly.com/media/indicator-closeup.mp4

A new close-up edit of the recorded roundabout sequence follows the fly pulling the indicator stalk, the Mini’s native bumper lamp flashing, and the stalk returning as the signal switches off.

The saved body motion and vehicle poses are replayed for this 60 fps slow-motion presentation. Navigation and speed remain engineered inputs. This is an inspection of the recorded control sequence, not a newly learned road-navigation result.

CREDITS
Project & images
The Driving Fly / Mark Unthank. Credit the project and link to thedrivingfly.com when covering these experiments. Sponsor logos remain the property of their owners.
https://thedrivingfly.com/

Measured connectivity
MaleCNS collaboration: HHMI Janelia FlyEM, University of Cambridge, MRC Laboratory of Molecular Biology and Google Research. Data: CC BY 4.0. The trained model uses engineered dynamics; its displayed activity is computed, not a recording of biological spikes.
https://male-cns.janelia.org/download/

Body & vehicle
NeuroMechFly / FlyGym (EPFL), Apache 2.0; CARLA 0.9.16 vehicle assets, CVC / Universitat Autònoma de Barcelona, CC BY. Sponsor surfaces and the cabin fly are composited for the films. No endorsement by these projects, MINI or BMW is implied.
https://thedrivingfly.com/media/attribution.md

Three-point-turn music
Boccherini, Minuet and Trio from String Quintet in E major, Op. 11 No. 5, G. 275. Recording: Rafael Krux, via FreePD / Wikimedia Commons; CC0 1.0. Excerpt with fades.
https://commons.wikimedia.org/wiki/File:Boccerini_Op11_n%C2%B05_G275_Satz3_Minuett.ogg

Parking music
Mozart, Eine kleine Nachtmusik, K. 525: I. Allegro. Musopen European Archive recording, identified as public domain by Wikimedia Commons; performing ensemble unspecified. Edited excerpt.
https://commons.wikimedia.org/wiki/File:Mozart_K525_Serenade_in_G_Major_1_-_Allegro.ogg

Horn recording
05 Horn.wav by 15HPanska_Ruttner_Jan, CC0 1.0. The recording is cropped, looped and faded to follow measured horn-button presses.
https://freesound.org/people/15HPanska_Ruttner_Jan/sounds/461679/
