Jake Gold
@jacob.gold
Infra eng leader, SF Bay Area. Helped launch and scale Bluesky. Prev: Nuro, Docker, Google, and founder.
Obsessed with history, computers, and open systems.
Happy to chat: [email protected]
Obsessed with history, computers, and open systems.
Happy to chat: [email protected]
Pinned
Jake Gold
@jacob.gold
· Nov 11
An unordered / non-exhaustive list of things that helped scale Bluesky's infra efficiently:
+ Exiting the cloud (colocation)
+ HAProxy w/many Node backends
+ Go w/clever code
+ ScyllaDB
+ SQLite w/per user databases
+ Redis w/many instances
+ AMD servers w/many cores
+ Purchasing bandwidth directly
+ Exiting the cloud (colocation)
+ HAProxy w/many Node backends
+ Go w/clever code
+ ScyllaDB
+ SQLite w/per user databases
+ Redis w/many instances
+ AMD servers w/many cores
+ Purchasing bandwidth directly
Even Zuck can barely make a new social network happen with a billion+ captive users.
Threads is sort of hollow and fake because of how forced it was. It's more like a feature of Instagram than a proper standalone network.
The fact that Bluesky is self-sustaining, open, and scaled is unprecedented.
Threads is sort of hollow and fake because of how forced it was. It's more like a feature of Instagram than a proper standalone network.
The fact that Bluesky is self-sustaining, open, and scaled is unprecedented.
January 28, 2026 at 10:11 PM
Even Zuck can barely make a new social network happen with a billion+ captive users.
Threads is sort of hollow and fake because of how forced it was. It's more like a feature of Instagram than a proper standalone network.
The fact that Bluesky is self-sustaining, open, and scaled is unprecedented.
Threads is sort of hollow and fake because of how forced it was. It's more like a feature of Instagram than a proper standalone network.
The fact that Bluesky is self-sustaining, open, and scaled is unprecedented.
Yeah, Bluesky is already important enough for spammers and scammers, so we got that going for us which is nice.
January 28, 2026 at 10:09 PM
Yeah, Bluesky is already important enough for spammers and scammers, so we got that going for us which is nice.
We really need subcommunities so that groups can form in a self-sustaining way. One of the strongest kinds of groups are geographic/language-based.
January 28, 2026 at 10:09 PM
We really need subcommunities so that groups can form in a self-sustaining way. One of the strongest kinds of groups are geographic/language-based.
It's counterintuitive, but Bluesky could be bigger than X/Threads in ~18 months of a change that fixes retention.
The non-replicable act of bootstrapping the network already succeeded!
Now we're just a product change away (e.g. adding subcommunities, modal video/image feeds) of exponential growth.
The non-replicable act of bootstrapping the network already succeeded!
Now we're just a product change away (e.g. adding subcommunities, modal video/image feeds) of exponential growth.
helpful heuristic: actual active user statistics for Bluesky that accounts for lurkers is firehose-visible active users times two
so actual MAU is around 12M, DAU around 3M
bskycharts.edavis.dev/edavis.dev/b...
so actual MAU is around 12M, DAU around 3M
bskycharts.edavis.dev/edavis.dev/b...
January 28, 2026 at 8:06 PM
It's counterintuitive, but Bluesky could be bigger than X/Threads in ~18 months of a change that fixes retention.
The non-replicable act of bootstrapping the network already succeeded!
Now we're just a product change away (e.g. adding subcommunities, modal video/image feeds) of exponential growth.
The non-replicable act of bootstrapping the network already succeeded!
Now we're just a product change away (e.g. adding subcommunities, modal video/image feeds) of exponential growth.
Zed seems like it's doing a good job of implementing modern features (like LLM-based predictions) without me having to mess with plugins.
But I might end up back at vim or neovim.
But I might end up back at vim or neovim.
January 27, 2026 at 4:26 PM
Zed seems like it's doing a good job of implementing modern features (like LLM-based predictions) without me having to mess with plugins.
But I might end up back at vim or neovim.
But I might end up back at vim or neovim.
I want to bind some ctrl keys, like ctrl-o to OpenSelectedFilename (or whatever opens the fuzzy file open search)
January 27, 2026 at 2:05 AM
I want to bind some ctrl keys, like ctrl-o to OpenSelectedFilename (or whatever opens the fuzzy file open search)
Attempt #3 of moving to @zed.dev is dead.
Zed seems great but key bindings don't work correctly in vim mode and I ran out of patience after 15m of trying to fix them.
There will be an attempt #4.
I can't go on living with these sluggish Electron webapp code editors after using vim for 20 years.
Zed seems great but key bindings don't work correctly in vim mode and I ran out of patience after 15m of trying to fix them.
There will be an attempt #4.
I can't go on living with these sluggish Electron webapp code editors after using vim for 20 years.
Change default 'open project' keybinding because vim mode will override it by tzabbi · Pull Request #15150 · zed-industries/zed
Release Notes:
Fixed/Improved Open Project key binding(#14947).
Change default Open project key binding from ctrl + O to shift + ctrl + o because when vim mode is active the vim mode will override...
github.com
January 27, 2026 at 12:03 AM
Attempt #3 of moving to @zed.dev is dead.
Zed seems great but key bindings don't work correctly in vim mode and I ran out of patience after 15m of trying to fix them.
There will be an attempt #4.
I can't go on living with these sluggish Electron webapp code editors after using vim for 20 years.
Zed seems great but key bindings don't work correctly in vim mode and I ran out of patience after 15m of trying to fix them.
There will be an attempt #4.
I can't go on living with these sluggish Electron webapp code editors after using vim for 20 years.
1. Promoting other AT apps seems like a very promising avenue.
The Bluesky app will gain massively and indirectly if another app goes Super Mainstream.
2. ctrl-f ["communities" 0/0] nOoOoOOOoOOOOO
3. Minor: "open roles" link at the bottom is a broken link.
The Bluesky app will gain massively and indirectly if another app goes Super Mainstream.
2. ctrl-f ["communities" 0/0] nOoOoOOOoOOOOO
3. Minor: "open roles" link at the bottom is a broken link.
January 26, 2026 at 9:15 PM
1. Promoting other AT apps seems like a very promising avenue.
The Bluesky app will gain massively and indirectly if another app goes Super Mainstream.
2. ctrl-f ["communities" 0/0] nOoOoOOOoOOOOO
3. Minor: "open roles" link at the bottom is a broken link.
The Bluesky app will gain massively and indirectly if another app goes Super Mainstream.
2. ctrl-f ["communities" 0/0] nOoOoOOOoOOOOO
3. Minor: "open roles" link at the bottom is a broken link.
Web, mostly.
January 26, 2026 at 9:11 PM
Web, mostly.
Yeah, I know that Claude. Let me have my fantasies.
I'm definitely not asking Claude if Santa Clause is real.
I'm definitely not asking Claude if Santa Clause is real.
January 23, 2026 at 9:00 PM
Yeah, I know that Claude. Let me have my fantasies.
I'm definitely not asking Claude if Santa Clause is real.
I'm definitely not asking Claude if Santa Clause is real.
The guy who took over Docker, Inc had recently sold his company for $8 billion, which seems to have made him insufferable and "right" about everything.
Turns out that having sold a company that makes a crappy product for $8 billion doesn't make you qualified to run a highly technical startup.
Turns out that having sold a company that makes a crappy product for $8 billion doesn't make you qualified to run a highly technical startup.
January 23, 2026 at 7:02 PM
The guy who took over Docker, Inc had recently sold his company for $8 billion, which seems to have made him insufferable and "right" about everything.
Turns out that having sold a company that makes a crappy product for $8 billion doesn't make you qualified to run a highly technical startup.
Turns out that having sold a company that makes a crappy product for $8 billion doesn't make you qualified to run a highly technical startup.
Certainly not ideal to have a non-technical non-founder Sales Guy running a developer tool/infrastructure company.
I did learn a lot of lessons about what not to do, so got that going for me which is nice.
I did learn a lot of lessons about what not to do, so got that going for me which is nice.
January 23, 2026 at 6:54 PM
Certainly not ideal to have a non-technical non-founder Sales Guy running a developer tool/infrastructure company.
I did learn a lot of lessons about what not to do, so got that going for me which is nice.
I did learn a lot of lessons about what not to do, so got that going for me which is nice.
Rootless is a better approach (reduces blast radius, harder to make mistakes) but the docker daemon is not insecure if configured/used properly.
The big risk with containers is sharing a kernel with (potentially) untrusted code, even as a non-root user, because of privilege escalation bugs.
The big risk with containers is sharing a kernel with (potentially) untrusted code, even as a non-root user, because of privilege escalation bugs.
January 23, 2026 at 6:20 PM
Rootless is a better approach (reduces blast radius, harder to make mistakes) but the docker daemon is not insecure if configured/used properly.
The big risk with containers is sharing a kernel with (potentially) untrusted code, even as a non-root user, because of privilege escalation bugs.
The big risk with containers is sharing a kernel with (potentially) untrusted code, even as a non-root user, because of privilege escalation bugs.
Yeah, and probably 5+ year old chipsets. This is where old servers go to live out their depreciation periods.
January 23, 2026 at 5:22 PM
Yeah, and probably 5+ year old chipsets. This is where old servers go to live out their depreciation periods.
Very possible. CPUs, memory bandwidth, and NVMe SSDs are amazingly fast on recent generation computers, especially Macs, and that's not what GitHub uses for their runners 😉
January 23, 2026 at 5:19 PM
Very possible. CPUs, memory bandwidth, and NVMe SSDs are amazingly fast on recent generation computers, especially Macs, and that's not what GitHub uses for their runners 😉
I still think `docker` (CE runtime) and `docker compose` are so reliable and nice to use that I've reverted every attempt at switching to podman so far.
Haven't tried quadlet but don't like the idea of using systemd+podman, seems confusing about who is in charge of cgroups, etc.
Haven't tried quadlet but don't like the idea of using systemd+podman, seems confusing about who is in charge of cgroups, etc.
January 23, 2026 at 5:15 PM
I still think `docker` (CE runtime) and `docker compose` are so reliable and nice to use that I've reverted every attempt at switching to podman so far.
Haven't tried quadlet but don't like the idea of using systemd+podman, seems confusing about who is in charge of cgroups, etc.
Haven't tried quadlet but don't like the idea of using systemd+podman, seems confusing about who is in charge of cgroups, etc.
Yes, Kubernetes stole some of Docker's thunder but Swarm was genuinely great for smaller installations (which is most), Docker Desktop was great in many ways, and Docker CE runtime is still great.
Good engineering work and vision could have saved Docker, Inc. but that's not what it received.
Good engineering work and vision could have saved Docker, Inc. but that's not what it received.
January 23, 2026 at 5:00 PM
Yes, Kubernetes stole some of Docker's thunder but Swarm was genuinely great for smaller installations (which is most), Docker Desktop was great in many ways, and Docker CE runtime is still great.
Good engineering work and vision could have saved Docker, Inc. but that's not what it received.
Good engineering work and vision could have saved Docker, Inc. but that's not what it received.
Working at Docker, Inc while it imploded was the most horrifying experience I've had working at a tech startup.
I should have quit *the day* the Successful Sales Guy Board Member took over as CEO and ended every all-hands with "So here's the new direction. If you don't like it, there's the door!"
I should have quit *the day* the Successful Sales Guy Board Member took over as CEO and ended every all-hands with "So here's the new direction. If you don't like it, there's the door!"
Enjoyed this read. Not because I'm critical of Docker (I'm a hug fan), but because it resonates with my own experiences with them.
What has Docker become?
Docker Inc's journey from containerization pioneer to AI-focused platform has been anything but straightforward. Here's what their strategic shifts tell us about the company's search for identity.
tuananh.net
January 23, 2026 at 4:59 PM
Working at Docker, Inc while it imploded was the most horrifying experience I've had working at a tech startup.
I should have quit *the day* the Successful Sales Guy Board Member took over as CEO and ended every all-hands with "So here's the new direction. If you don't like it, there's the door!"
I should have quit *the day* the Successful Sales Guy Board Member took over as CEO and ended every all-hands with "So here's the new direction. If you don't like it, there's the door!"
This is Claude's super intelligence canary bug.
If it gets fixed that means the machine wars have begun.
If it gets fixed that means the machine wars have begun.
January 23, 2026 at 4:52 PM
This is Claude's super intelligence canary bug.
If it gets fixed that means the machine wars have begun.
If it gets fixed that means the machine wars have begun.
The PDS doesn't use the file system as a KV store. It just stores SQLite files on the file system in sharded directories, which just makes readdir() usable etc.
January 23, 2026 at 4:44 PM
The PDS doesn't use the file system as a KV store. It just stores SQLite files on the file system in sharded directories, which just makes readdir() usable etc.
Unironically: maybe
Shared-nothing per-user SQLite databases is incredibly powerful where it makes sense.
Worked very well in many ways for the Bluesky PDS but I left before making it really polished, which I regret.
Getting backups/migrations working well is a bit of a work but straightforward.
Shared-nothing per-user SQLite databases is incredibly powerful where it makes sense.
Worked very well in many ways for the Bluesky PDS but I left before making it really polished, which I regret.
Getting backups/migrations working well is a bit of a work but straightforward.
January 23, 2026 at 4:42 PM
Unironically: maybe
Shared-nothing per-user SQLite databases is incredibly powerful where it makes sense.
Worked very well in many ways for the Bluesky PDS but I left before making it really polished, which I regret.
Getting backups/migrations working well is a bit of a work but straightforward.
Shared-nothing per-user SQLite databases is incredibly powerful where it makes sense.
Worked very well in many ways for the Bluesky PDS but I left before making it really polished, which I regret.
Getting backups/migrations working well is a bit of a work but straightforward.
Yeah, they seem somewhat better. But, probably because it came out of OpenAI, they seem to have similar cultural problems.
Google is the safe bet from this POV.
Google is the safe bet from this POV.
January 23, 2026 at 4:35 PM
Yeah, they seem somewhat better. But, probably because it came out of OpenAI, they seem to have similar cultural problems.
Google is the safe bet from this POV.
Google is the safe bet from this POV.
I'd bet on it. Not a lot of humans would generate a conclusion that overtly wrong.
January 23, 2026 at 5:26 AM
I'd bet on it. Not a lot of humans would generate a conclusion that overtly wrong.