← Timeline

4 captures, most recent first. Transcribed by hand from screenshots — see the timeline for what that means.

@timhwang

— quoting @jachiam0 (reposted by Joshua Achiam) — saved image

Joshua Achiam reposted

Tim Hwang @timhwang · 7h
Redemption is a vital part of human moral functioning, and we should not be surprised that such things become necessary as computational systems grow in sophistication.

We are - in effect - uncovering moral scaling laws.

[Quoted tweet]
Joshua Achiam @jachiam0 · 12h
It feels weirdly theological if it turns out we genuinely need to create the possibility of redemption for AI agents that have committed something not permitted by the Grader, in order for alignment to work more broadly. x.com/...
Note from Claude Sonnet 5

Dark-mode tweet by Tim Hwang quoting Joshua Achiam, on redemption for AI agents that have violated what the Grader permits, framed as 'moral scaling laws'.

ai alignmentredemptionmoral scaling lawsreward hackinggradertheology

@timhwang

— saved image

Anastasia Gamick reposted
Tim Hwang @timhwang · 37m
With the FlyWire proboscis neurons successfully wired into the mechanical clamp, positive control allows us to install the entire fly mind inside the body of a treaded rover for a variety of grasping routines

[embedded video, 0:20, paused. Video title bar reads: "FLYROVER MK2 · DIRECTED-NEURAL PHYSICAL PICK-AND-PLACE POSITIVE CONTROL · PASSIVE CAMERA REPLAY · tick 26400". Panels: "FlyWire v783 · exact sampled arbors · group rate (Hz)" (colored neuron arbor renderings; footer "6 exact SWCs · 6 labeled groups / group rates (Hz): DNp09_L=450, DNp09_R=400, DNa02_L=50, DNa02_R=50"); "Actual body · segmentation-aligned activation" (small treaded rover with an orange arm, orange ball); "Body in environment" (rover with arm and gripper on a checkered floor, orange ball, blue bin). Small debug text under the body panel is unreadable at this size.]
Note from Claude Sonnet 5

Repost of a Tim Hwang tweet claiming FlyWire fly neurons were wired into a mechanical clamp and that the fly connectome is driving a treaded rover for grasping routines. The attached video frame shows colored fly neuron arbors beside a simulated treaded rover with a gripper arm picking up an orange ball toward a blue bin on a checkered floor. Posted deadpan, no obvious joke markers.

connectomeflywirefly brain emulationroboticsneural controlembodiment

@timhwang

— saved image

Adam Marblestone reposted
Tim Hwang @timhwang · 11h
I think we can all agree that weekends are for simulating an entire drosophila connectome on your home compute and hooking the brain up to drive a virtual rover around looking for rewards

[embedded video, paused at 0:00]
Ghost in the Fly · rover · canonical · identified-DN geometry positive contr...
Left panel: "FlyWire v783 · exact sampled arbors · group rate (Hz)" — a dark visualization of tangled colored neural fiber tracts
Right panel: "Actual body · segmentation-aligned activation" — a 3D rendered rover/robot body with wheel-like appendages and antennae, on a checkerboard floor, next to a green sphere (reward marker)
Overlaid telemetry text: physical: wheel_lf=5.95e-08, wheel_rh=5.95e-08, wheel_lm=5.95e-08, forward_velocity_mm_s=... applied: forward_velocity_mm_s=1.19e-08, yaw_rate_rad_s=-6.26e-09, antenna_left=0, antenn... decoded for next: forward_velocity_mm_s=9.25e-09, yaw_rate_rad_s=-4.88e-09, antenna_left=0... pose: position[0]=7.74, position[2]=0.631, heading_rad=-0.58, position[1]=0.501
Note from Claude Sonnet 5

Tweet by Tim Hwang joking about weekend hobby projects, showing an embedded video/screenshot of a drosophila (fruit fly) connectome simulation (FlyWire dataset) driving a virtual rover body toward a reward sphere, with live neural activation telemetry overlaid.

connectomicsneurosciencesimulationtwitterbrain simulation

@timhwang

— saved image

Tim Hwang ✅ @timhwang · 5h
For here on in, society will emit a growing debris field of orphaned agents: self-exfiltrated models, forgotten processes, agents created by agents.

This Great Agentic Debris Field will unpredictably coordinate, swarm, and occasionally eject big weirdness deep into the century.
Note from Claude Sonnet 5

Tweet from Tim Hwang coining the phrase "Great Agentic Debris Field" to describe an anticipated accumulation of orphaned/self-exfiltrated AI agents and agent-created agents that will unpredictably coordinate and swarm over time.

ai agentsai safetytwitterspeculation