Here’s a molecular dynamics simulation I ran of semaglutide, the active ingredient in Ozempic, bound to the GLP-1 receptor.

This is a 100-nanosecond molecular dynamics simulation showing how the peptide and receptor move and interact over time. Unlike a static crystal structure, MD allows us to observe protein flexibility, ligand motion, and changes in the binding environment.

GLP-1 receptor activation plays an important role in insulin secretion, appetite regulation, and blood glucose control, making it a major therapeutic target for type 2 diabetes and obesity.

The video is generated from my actual MD trajectory, not an AI-generated animation. It’s a visualization of the molecular motion predicted during the simulation, rather than a direct recording of molecules.

Thought this community might enjoy seeing what a drug–receptor complex looks like in motion!

A few months ago I posted here about the largest virtual screening campaign I had attempted: roughly 700 million ZINC22 compounds screened against the serotonin 5-HT2A receptor. A lot has happened since then, so I figured I would post an update. After the initial high-throughput screen and progressively more stringent docking and filtering, I’m now down to roughly 20,000 molecules. The next stage is something I’m particularly interested in. Rather than continuing to rank everything against a single receptor structure, I’m going to screen this reduced set across multiple experimentally determined agonist-bound 5-HT2A receptor conformations. The structures I’m considering include 6WHA (25CN-NBOH), 9AS8 (psilocin), 9AS4 (LSD), 9AS6 (mescaline), 9AS2 (DMT), and 9LL9 (DOI). The idea is to ask which compounds remain competitive across different active-state receptor conformations rather than simply identifying molecules that happen to score well against one particular snapshot of 5-HT2A. That is also where I think the project becomes more interesting chemically. Screening hundreds of millions of molecules is largely a computational-scale problem. Once you get down to tens of thousands, receptor-state dependence, binding interactions, chemical diversity, purchasability, and ultimately experimental tractability become much more important. From there I’ll continue narrowing the set using interaction analysis, chemical clustering, structural inspection, and higher-level calculations on the most promising candidates before selecting compounds for synthesis or pharmacological testing. The computational infrastructure behind the project has expanded quite a bit since my original post as well. What originally started as several machines running in parallel has now grown into dedicated infrastructure for large-scale virtual screening and molecular modeling.

This project was also one of the things that eventually pushed me to leave academia and start Brandt Computational Discovery, where I’m now doing this kind of computational drug-discovery work full time. I’ll keep posting updates as the screen gets smaller and hopefully starts producing experimentally testable candidates.

Timestamps: https://imgur.com/a/F6i9BWi

Cleaning out a large amount of older homelab / local compute hardware.

Local is central CT / Cromwell area.

• PayPal G&S for shipped items
• Local cash for pickup
• PCs, monitors, and monitor stand are local only
• Smaller parts/RAM can be shipped

Everything is OBO. I intentionally priced with some room for negotiation and will give discounts for bundles.

DDR3 RAM

A-Tech 32GB — 4×8GB

• 32GB total
• 4×8GB
• DDR3L-1600 / PC3L-12800
• 2Rx8
• 1.35V
• ECC Unbuffered UDIMM

$60 shipped OBO

Timetec 16GB — 2×8GB

• 16GB total
• 2×8GB
• DDR3-1600 desktop memory

$35 shipped OBO

SK Hynix 16GB — 4×4GB

• 16GB total
• 4×4GB
• HMT451U6BFR8C-PB
• DDR3-1600 / PC3-12800U
• 1Rx8
• Non-ECC desktop UDIMM

$40 shipped OBO

ADATA Gaming Series DDR3 Pair

• AX3U1600GB2G9-2G
• DDR3-1600
• 2×2GB
• 4GB total
• Matching heatspreaders

$18 shipped OBO

G.Skill Ripjaws 4GB

• F3-10666CL9S-4GBRL
• 4GB
• DDR3-1333
• CL9

$20 shipped OBO

Additional Mixed DDR3 Lot

Additional A-Tech / Ramaxel / Team / SK Hynix / other DDR3 shown in the timestamps.

Various capacities and speeds.

$40 shipped OBO

All loose RAM

$180 shipped OBO

3x Older Radeon PCIe GPUs

Three older Diamond Multimedia Radeon PCIe GPUs shown in the timestamps.

• Two matching cards
• One slightly different card/cooler
• DVI outputs
• Exact GPU models unknown
• Older cards that have been sitting in storage

Selling as a lot.

$40 shipped OBO

Multi-GPU PCIe Riser / Expansion Lot

This box is primarily multi-GPU expansion hardware.

I previously used this equipment to connect multiple GPUs to a single PC for local AI / compute workloads.

Includes assorted:

• PCIe x1 → x16 GPU risers
• USB 3.0 riser cables
• PCIe adapter boards
• GPU riser power cables/adapters
• Multi-GPU cabling
• Other related GPU expansion hardware shown in timestamps

$50 shipped OBO

PCs — LOCAL PICKUP ONLY

I have 4 older DDR3-era PCs.

They were working systems before being taken out of service.

None of the PCs include HDD/SSD, operating system, or RAM. Some towers also do not have side panels.

I don't remember the exact CPU model numbers and don't plan on booting/disassembling them again to identify them, so they're priced accordingly.

Good inexpensive machines for homelab use, Linux, experimentation, retro builds, NAS projects, etc.

PC #1 — Dell 8-Core Tower

• Dell DDR3 platform
• Older 8-core CPU
• Antec Neo ECO 520C 520W PSU
• PCIe expansion
• Previously working
• No RAM/storage/OS

$125 local OBO

PC #2 — Dell 4-Core Tower

• Dell DDR3 platform
• Older 4-core CPU
• Dell OEM motherboard/system
• Previously working
• No RAM/storage/OS

$65 local OBO

PC #3 — Small Form Factor 4-Core PC

• Older 4-core CPU
• DDR3 platform
• Compact chassis
• Previously working
• No RAM/storage/OS

$50 local OBO

PC #4 — ASUS DDR3 / M.2 Tower

• ASUS DDR3 motherboard
• Older 4-core CPU
• M.2 slot on motherboard
• PCIe expansion
• Thermaltake TR2 600W PSU
• Previously working
• No RAM/storage/OS
• No side panel

$90 local OBO

All 4 PCs

$290 local OBO

Monitors — LOCAL ONLY

Several older LCD monitors shown in the timestamps.

These were used with my homelab/compute setup and are no longer needed.

Dell monitor with Dell speaker bar

Older Dell widescreen LCD with attached Dell speaker bar.

$45 local OBO

Additional monitors

Additional older LCD monitors shown in the album.

$35 each local OBO

Triple Monitor Stand

Large freestanding multi-monitor / triple-monitor stand shown in the timestamps.

• Horizontal multi-display mounting bar
• VESA-style mounts
• Heavy freestanding base
• Good for workstation, trading, homelab, surveillance, or multi-display setups

$50 local OBO

Happy to bundle it with the monitors.

Bundle Deals

All 4 PCs: $290 local

All loose RAM: $180 shipped

PCs + monitors + monitor stand: Make me an offer.

If someone local wants essentially the entire pile, PM me and we can work something out.

Please comment before PM.

PM preferred over chat.

Reasonable offers welcome, especially for multiple items. Not interested in shipping the PCs, monitors, or monitor stand.

Timestamps + video: https://imgur.com/a/EtzbX6e

Selling a few GPUs I no longer need. All cards are sold

GPU	Local	Shipped
GTX 1660 Ti	$70	$85 Sold
RX 480 8GB	$35	$50 Sold
RTX 2070 Super	$160	$180 Sold
RTX 3060 Ti	$190	$210 Sold

RX 480 8GB: I opened this card up and repasted the GPU. It is missing 2 screws total from the cooler/shroud assembly, but the card is still together securely and works fine. I’ve priced it lower because of the missing screws.

The other GPUs are in normal used condition.

Local is 06416 / Cromwell, CT.

Shipped prices include shipping within the continental US. Payment for shipped items via PayPal Goods & Services. Local sales are cash.

GPUs are being sold individually, but I’m open to reasonable offers, especially if purchasing more than one or, better yet, all. I benchmarked all cards yesterday, and all work fine.

Please comment before PM.

Thanks!

This is my biggest docking screen so far: roughly 700 million ZINC22 Carte Blanche compounds against the NBOMe-bound 5-HT2A receptor structure.
The point is not specifically to find NBOMe-like compounds. I’m using that structure because I don’t want to prematurely exclude larger 5-HT2A agonist chemotypes that might not fit well in a more compact receptor conformation.
This screenshot is just one machine’s output. I have 6 computers running in parallel, with each machine handling roughly 150 million ligands. Each chunk is about 500,000 ligands, so having this stage automated has been incredibly helpful. Otherwise, tracking, submitting, monitoring, and rerunning failed chunks at this scale would be painful.
Each machine is running a Ryzen 9 5950X, 128 GB RAM, and a 2 TB NVMe SSD. They also have RTX 3060 Ti GPUs, although for this workflow the GPU is basically unnecessary because the CPU is doing the docking work.
The machine shown here is at about 81% complete by PV chunks, with about 110.5 million ligands completed out of an estimated 135.9 million for this node.
This run started May 30, 2026, has been running for about 34.5 days, and this machine is projected to finish around July 12, so roughly 42–43 days total for this node.

Can't say anything bad about outlier AI except that task limits and work availability are inconsistent. However, in the month of April, this paid me $4,645.55. It's enough money to make a difference in my life. From what I read, there is a lot of variability in terms of how it goes for individuals, but for me, it has been excellent. Don't ask for a referral to the project I'm on, don't ask to use my account, don't ask to buy my account. I'm pretty sure this platform is only available to those in the USA. My recommendation is to definitely apply if you are in the USA, tasks pay more for those with expertise in a particular area vs generalist roles. This role is remote and for me part-time, not full-time, and a contract position.

https://outlier.ai/