Sign in

usrbinkat.io

@usrbinkat.io
4.5K followers 5.4K following 4.7K posts

🧙‍♀️ Open Sourceress 🚀 Ms. (f)Rizzle @Cisco 🏳️‍⚧️ Blame for Oops & Ops my own 🤗 Seize the means of computing 👩‍💻 K8s/AWS/Azure Platform Engineer 📖 blog.usrbinkat.io/en/page/about 🍿Watching Dystopia IRL 🧠 Neuro Spicy Autist 📍 Sacramento ☀

PostsRepliesMedia
usrbinkat.io @usrbinkat.io · 14/09/2026
packaged opencode2 cli and desktop in flake.nix, the same again for codex. both were very crude to start with. now opencode-desktop2 + astra is running all contributing.md static analysis after building and testing. reading between the lines it's def worth watching.
050
usrbinkat.io @usrbinkat.io · 13/09/2026
Dear reader, the plural of corn is corn. 🌽 🌽 Corns are, and I don't want to hear it otherwise, corns are growths on feet. Please do not purchase corns at the grocery store. I think we should invest more in non artificial intelligence. 🙏
060
usrbinkat.io @usrbinkat.io · 13/09/2026
Chat. They are selling Corns. 😭😭
370
usrbinkat.io @usrbinkat.io · 18/08/2026
OMG it's pulling from an OG dis track.
060
usrbinkat.io @usrbinkat.io · 10/08/2026
Who put the Loctite on the talent lid wtf 😭
280
usrbinkat.io @usrbinkat.io · 07/08/2026
I just discovered Lucas Lenwas omg So much flavor and it's sweet and a little spicy and balanced with a little savory. It was almost overwhelming.
050
usrbinkat.io @usrbinkat.io · 07/08/2026
We just road tripped to Reno Nevada! 🛻
120
usrbinkat.io @usrbinkat.io · 28/07/2026
But the 92 heritage edition had the Corvette L98 Tuned Port Injection from the factory! *sniffles in ancient history
120
usrbinkat.io @usrbinkat.io · 23/07/2026
This is actually a life goal, I'm low key (high key) dumbfounded. (rust oci conformant registry)
080
usrbinkat.io @usrbinkat.io · 18/07/2026
Look ma' OpenSesame updated for COSMIC 1.3.0!!
060
usrbinkat.io @usrbinkat.io · 15/07/2026
#neversurrender
040
usrbinkat.io @usrbinkat.io · 15/07/2026
No I will not calm down now.
2192
usrbinkat.io @usrbinkat.io · 11/07/2026
070
usrbinkat.io @usrbinkat.io · 25/06/2026
No good upgrade is complete without a benchmark ✨ Sweet dreams are made of these 💤
A screenshot shows terminal output of a Ceph Storage Benchmark running on the NVMe tier from node op1 in a tmux session. The benchmark configuration header shows pod bench-nvme running tests for seqwrite, seqread, randwrite, randread, and mixed workloads with 25G file size, IO depth 32, 4 jobs, 30s runtime with time-based tests, using O_DIRECT, mounted at /data. The benchmark date is 2026-06-25.

Five test results display in sequence. Sequential Write achieves 963.09 MB/s at 963 IOPS with 132.882ms latency. Sequential Read reaches 1796.80 MB/s at 1797 IOPS with 71.199ms latency. Random 4K Write delivers 46.13 MB/s at 11812 IOPS with 10.834ms latency. Random 4K Read hits 191.25 MB/s at 48961 IOPS with 2.612ms latency. Mixed 70/30 4K workload shows 34.43 MB/s reads at 8816 IOPS with 3.919ms latency and 34.43 MB/s writes at 8816 IOPS with 10.595ms latency. Below the results, a formatted markdown Ceph Storage Benchmark Report begins with parameter and results tables.

The benchmark validates the replacement WD_BLACK SN850X drives are performing well in the rebuilt Ceph cluster, with nearly 1.8 GB/s sequential reads and almost 49K random read IOPS through the full Ceph stack including network replication. These numbers reflect real distributed storage performance rather than raw drive speeds, with the sequential write throughput around 963 MB/s indicating the overhead of Ceph's replication factor of 2 across the network while random 4K performance demonstrates the SN850X's strong small-block capability translating effectively through the Rook-Ceph CSI layer on Talos Linux.
150
usrbinkat.io @usrbinkat.io · 25/06/2026
Storage class is in session.
A screenshot shows K9s terminal UI displaying 16 Kubernetes StorageClasses on the optiplexprime-konduktor-admin cluster running Kubernetes 1.36.2. A keyboard shortcut menu is visible in the header showing options for Delete, Describe, Edit, Help, Jump Owner, and YAML. The storageclass view lists resources in a table with columns for Name, Provisioner, ReclaimPolicy, and VolumeBindingMode.

The storage classes span four Ceph storage backends. S3 bucket classes include ceph-bucket-s3-archive highlighted in blue, ceph-bucket-s3-nvme, and ceph-bucket-s3-ssd, all using rook-ceph.ceph.rook.io/bucket provisioner. NVMeoF classes include ceph-nvmeof-hdd, ceph-nvmeof-nvme, and ceph-nvmeof-ssd using rook-ceph.nvmeof.csi.ceph.com with Immediate binding. RBD block classes appear in pairs with Delete and Retain reclaim policies for hdd, nvme, and ssd tiers using rook-ceph.rbd.csi.ceph.com, with rook-ceph-block-nvme marked as cluster default. CephFS classes include rook-cephfs-hdd, rook-cephfs-nvme, and rook-cephfs-ssd using rook-ceph.cephfs.csi.ceph.com. A hostpath-provisioner for KubeVirt also appears. All non-NVMeoF classes use WaitForFirstConsumer binding.

This comprehensive storage class matrix demonstrates the full tiered storage architecture now operational after the NVMe drive replacement and Pulumi deployment. The NVMe tier set as cluster default reflects prioritizing the new WD_BLACK SN850X drives for general workloads, while the systematic coverage of block, filesystem, object, and NVMeoF access modes across HDD, SSD, and NVMe device classes provides Kubernetes workloads with granular performance selection entirely managed through declarative IaC.
160
usrbinkat.io @usrbinkat.io · 25/06/2026
That was actually **too easy** I love it when a plan comes together.
A screenshot shows K9s terminal UI displaying Ceph block pools in the rook-ceph namespace on Kubernetes 1.36.2. The K9s header shows context usrbinkat-optiplexprime-konduktor-admin in RW mode with K9s version 0.51.0, and a namespace selector listing all, authelia-system, default, rook-ceph, and braincraft. The ASCII K9s dog logo appears in the upper right corner.

The cephblockpools(all)[4] view lists four block pools in a table with columns for Namespace, Name, Phase, Type, Failuredomain, and Age. All four pools show rook-ceph namespace, Ready phase, Replicated type, and host failure domain. The pools are hdd-blockpool at 28h, nvme-blockpool highlighted in blue at 2m59s indicating it was just created, nvmeof at 26h, and ssd-blockpool at 28h. The bottom bar shows navigation breadcrumbs for customresourcedefinition and cephblockpool.

The freshly created nvme-blockpool at just under 3 minutes old confirms the Pulumi deployment successfully provisioned the new NVMe-tier block storage on the replacement WD_BLACK drives. The four-tier block pool architecture spanning HDD, SSD, NVMe, and NVMeoF device classes demonstrates granular storage tiering within a single Ceph cluster, enabling workload-appropriate performance selection through Kubernetes StorageClasses all managed declaratively through Pulumi IaC.
130
usrbinkat.io @usrbinkat.io · 25/06/2026
***always makes me glimmer***
A screenshot shows Pulumi executing a stack update for optiplex-admin-storage in a tmux terminal session. The command pulumi up --stack optiplex-admin-storage --skip-preview=true --refresh=true runs from usrbinkat@mithril connected to the optiplexprime-konduktor-admin:default braincraft main context. A browser link points to app.pulumi.com/usrbinkat/infrastructure/optiplex-admin-storage/updates/26.

The output displays a hierarchical resource tree with columns for Type, Name, Status, and Info. The tree shows the pulumi:pulumi:Stack at the root refreshing with a resource warning, branching into dozens of Kubernetes resources. Visible components include k8s-provider, zot-registry chart with its ServiceAccount, ConfigMap, Service, and Secret, snapshot-controller with RBAC and deployment resources, cloudnative-pg-operator ConfigMap, redis-operator chart with deployment and RBAC bindings, and an extensive rook-ceph-operator-chart section dominating the output with dozens of ServiceAccounts, Roles, ClusterRoles, ClusterRoleBindings, CustomResourceDefinitions, and RoleBindings for ceph-csi, rook-ceph-osd, rook-ceph-mgr, rook-ceph-system, objectstorage-provisioner, and other Ceph components. Status column shows refresh timings ranging from 0s to 1s. A kubeconfig-store dynamic resource appears at the bottom.

This captures the Pulumi IaC deployment applying the updated storage configuration with the new WD_BLACK NVMe device definitions and NVMe-tier CephFS and S3 configurations. The extensive Rook Ceph operator resource tree reflects the complexity of declaratively managing distributed storage through pure infrastructure as code, with every Kubernetes resource from RBAC policies to CRDs version-controlled and deployed programmatically without a single helm or kubectl command.
120
usrbinkat.io @usrbinkat.io · 25/06/2026
"mom can i buy some s3?" "honey we have s3 at home"
A screenshot shows Claude Code in a tmux terminal session displaying a YAML diff adding an NVMe-tier RADOS Gateway S3 object storage configuration to Pulumi.optiplex-admin-storage.yaml. Green highlighted lines from 546 to 569 define the new block commented as NVMe tier - Hot data with 1TB WD SN850X TLC. The configuration specifies enabled true, name s3-nvme, device_class nvme, instances 1, port 80, metadata_pool_replicated_size 3, data_pool_replicated_size 2, data_pool_erasure_coded false, and resource limits of 250m/1 CPU and 512Mi/2Gi memory. The route section enables an httproute with hostname s3.nvme.optiplexprime.home.arpa, additional hostname s3.nvme.braincraft.io, gateway_type envoy, gateway_name envoy-gateway in envoy-gateway-system namespace, and gateway_listener_name https. Below, the existing SSD tier S3 configuration with s3-ssd remains visible. Claude Code prompts with option 1 Yes selected.

This adds a dedicated NVMe-backed S3-compatible object storage endpoint alongside the existing SSD tier, creating a hot-data storage class on the replacement WD_BLACK drives. The Envoy gateway routing with both local .home.arpa and public braincraft.io hostnames enables S3 API access from both the homelab network and external services. The dual-tier S3 architecture separates hot and warm data across NVMe and SSD device classes, with metadata at triple replication for durability and data at double replication to balance redundancy with the 1TB per-node NVMe capacity.
120
usrbinkat.io @usrbinkat.io · 25/06/2026
So tempted to turn replication off to get that raw speeeeed... But if I wanted speed I'd find a sugar to buy me ConnectX-6 Lx. Girl's gotta choose between shoes and the rack. Inconceivable.
A screenshot shows Claude Code in a tmux terminal session displaying a YAML diff adding a new NVMe-tier CephFS configuration to Pulumi.optiplex-admin-storage.yaml. Green highlighted lines from 475 to 490 show the new block being added with comment NVMe tier CephFS 1TB WD SN850X TLC. The configuration defines enabled true, name rook-cephfs-nvme, filesystem name cephfs-nvme, pool name data0, device class nvme, is_default false, replicated_size 2, metadata_replicated_size 3, reclaim_policy Delete, allow_volume_expansion true, volume_binding_mode WaitForFirstConsumer, mds_active_count 1, mds_standby_count 1, and pg_autoscale_mode_data and metadata both set to on. Below the new block, the existing SSD tier CephFS configuration beginning with rook-cephfs-ssd remains visible for context. Claude Code prompts with option 1 Yes selected.

This adds a dedicated NVMe-backed CephFS filesystem alongside the existing SSD tier, creating a high-performance storage class on the newly installed WD_BLACK SN850X drives. CephFS provides ReadWriteMany access mode essential for KubeVirt live migration, where virtual machine disks must be simultaneously accessible from both source and destination nodes during migration. The replicated_size of 2 balances redundancy with capacity across the three-node cluster, while WaitForFirstConsumer binding ensures volumes are provisioned on nodes with available NVMe-class OSDs.
130
usrbinkat.io @usrbinkat.io · 25/06/2026
That dialed in Pulumi IaC makes this unreasonably easy.
A screenshot shows Claude Code in a tmux terminal session displaying a YAML diff for the file Pulumi.optiplex-admin-storage.yaml. The diff shows changes across three Ceph storage nodes: op1 at 10.0.0.33, op2 at 10.0.0.34, and op3 at 10.0.0.35. For each node, red highlighted lines show commented-out old Sabrent NVMe devices marked as offline for maintenance, while green highlighted lines show their replacements: new WD_BLACK SN850X 1TB TLC drives with full /dev/disk/by-id paths. Each node retains an unchanged secondary SSD device alongside the replaced NVMe.

The device paths show the naming convention shift from nvme-Sabrent identifiers to nvme-WD_BLACK_SN850X_HS_1000GB with unique serial numbers for each of the three drives. Node op1 includes an additional comment noting the 2TB Sabrent was replaced and is now the Talos system install disk. At the bottom, Claude Code prompts Do you want to make this edit to Pulumi.optiplex-admin-storage.yaml? with option 2 selected: Yes, allow all edits during this session.

This captures the infrastructure-as-code approach to Ceph cluster recovery using Pulumi Python IaC, where replacing failed drives means updating YAML configuration with new device identifiers rather than running manual kubectl or ceph commands. Claude Code is assisting with the declarative storage infrastructure update, embodying the no-helm-no-kubectl philosophy where all cluster state is managed through version-controlled Pulumi definitions deployed programmatically.
240
usrbinkat.io @usrbinkat.io · 25/06/2026
Talos showing the new block devices on all nodes :D Thank you @siderolabs.com
A screenshot shows the Sidero Omni web interface displaying the Disks tab for node op1 in a dark-themed dashboard. The interface shows two NVMe drives. The first, nvme0n1 at /dev/nvme0n1, is a 1.0 TB nvme drive showing entirely as Unallocated 931.51 GB in gray, representing the newly installed replacement WD_BLACK SN850X ready for Ceph OSD provisioning.

The second drive, nvme1n1 at /dev/nvme1n1, is a 2.0 TB nvme drive with a horizontal bar visualization showing partition layout. Four partitions are listed below: EFI at /dev/nvme1n1p1 with vfat filesystem, unknown encryption, and 2.2 GB size; META at /dev/nvme1n1p2 with talosmeta filesystem, encryption disabled, at 1.0 MB; STATE at /dev/nvme1n1p3 with xfs filesystem, encryption disabled, with UUID shown, at 105 MB; and EPHEMERAL at /dev/nvme1n1p4 with xfs filesystem, encryption disabled, at 2.0 TB shown as the dominant blue bar in the visualization. A color-coded legend identifies each partition type.

The Sidero Omni management plane confirms the replacement NVMe is detected and unallocated on Talos Linux node op1, ready for Ceph to claim as a new OSD. The existing 2 TB boot drive shows Talos's characteristic minimal partition scheme with tiny EFI, META, and STATE partitions alongside the large ephemeral partition, reflecting Talos's immutable OS design where the system runs from a read-only image and only ephemeral workload data persists.
1150
usrbinkat.io @usrbinkat.io · 25/06/2026
I did the maintenance on one node at a time. Is this... maturity?
A photograph under magenta LED lighting shows the rear view of multiple Dell small form factor workstations stacked vertically, their back panels dense with cables connected to a white network switch with SFP+ ports visible across the top. Ethernet cables in red, white, and gray run between the nodes and switch, while blue and green LED indicators glow through the tangle of power and data cables. Each chassis shows ventilation grilles, USB ports, HDMI outputs, and DisplayPort connections. A sticker and other desk items are partially visible on the shelf above.

The three Ceph storage nodes with freshly installed WD_BLACK NVMe drives are wired back into the homelab cluster and powering on, connected via the SFP+ capable switch for the high-bandwidth interconnect that distributed storage demands. The beautiful chaos of cables reflects the reality of homelab infrastructure where enterprise-grade Kubernetes and Ceph deployments live alongside desk clutter, with fingers crossed that the OSDs rebuild cleanly and placement groups return to active plus clean status.
1110
usrbinkat.io @usrbinkat.io · 25/06/2026
*slaps hood of optiplex* yOu CaN fIt SoMaNy TBs iN tH Is bAd bOY
A photograph bathed in magenta lighting shows a hand inserting a WD_BLACK NVMe SSD into an open Dell small form factor workstation. The compact chassis reveals densely packed internals including a second NVMe already seated in an M.2 slot, SATA cables, an SK hynix branded component, and a traditional spinning hard drive visible in the drive bay. The Dell branding is partially visible on the chassis edge.

This captures the hands-on hardware work of installing replacement NVMe drives into one of three Ceph storage cluster nodes after a drive failure took out the distributed storage system. The small form factor Dell machines serve as compact Kubernetes nodes running Talos Linux, packing enterprise storage capability into desktop-class hardware for a homelab Ceph deployment providing CephFS, RBD block, and S3 object storage.
1110
usrbinkat.io @usrbinkat.io · 25/06/2026
Rook Ceph killed a humble OEM nvme storage pool in my second hand optiplex cluster. This evening's project is rebuilding the NVME tier storage class. Pray to the old gods and the new🐧
Three WD_BLACK SN850X NVMe SSDs fanned out on a grid cutting mat under vivid magenta LED lighting. Two drives show the black heatsink label with white branding, the third is flipped revealing its green PCB and gold M.2 connector pins. A white mechanical keyboard with green backlighting sits in the upper left, a grapefruit tequila seltzer can in the upper right, and a white desk lamp base along the right edge.

Replacement drives staged for installation into three Ceph storage cluster nodes after a failed NVMe cascaded into a cluster outage. The SN850X's PCIe Gen4 speeds suit the distributed storage workload across Talos Linux and Kubernetes. The tequila seltzer and aesthetic purple lighting capture the vibe of late-night infrastructure recovery, hardware prep before seating drives, rebuilding OSDs, and waiting for Ceph to rebalance.
2491
usrbinkat.io @usrbinkat.io · 19/06/2026
I wake up. I have coffee. I read a short FFmpeg blog. There are no short FFmpeg blogs.
2121
usrbinkat.io @usrbinkat.io · 17/06/2026
I was just reading this from anthropic this morning and I cannot fathom letting Claude pull design and implementation decision making out of its ass. The two constants with Claude are that in the face of knowing what is correct, it will always downgrade from correct to easier, easiest, simplest.
https://www.anthropic.com/research/claude-code-expertise
100
usrbinkat.io @usrbinkat.io · 16/06/2026
I'm dizzy from IO_URING, rayon, tokio 😵‍💫 What I did discover, WRITING SOFTWARE IS HARD WORK (even with AI). What I did NOT expect to discover, was how related cloud scale microservice engineering is to writing local programs.
0150
usrbinkat.io @usrbinkat.io · 23/05/2026
i want to be excited about space but cmon
060
usrbinkat.io @usrbinkat.io · 23/05/2026
That scroll do be dooming
1130
usrbinkat.io @usrbinkat.io · 21/05/2026
Hey, creative technologists, this is cool.
0354
usrbinkat.io @usrbinkat.io · 21/05/2026
I am unreasonably excited for this. Thank you for sharing like omg!
180
usrbinkat.io @usrbinkat.io · 21/05/2026
I've been developing on more modest hardware for the last few weeks which has inadvertently but appropriately forced more responsible system resource cost criteria. So that's great right? right? anyway no lessons learned.. 256GB nvme is terribly smol. I am spoiled how does anyone live like this.
8280
usrbinkat.io @usrbinkat.io · 19/05/2026
Also I'm sorry I build the Internet. I know it's completely unusable. @gizmodo.com your website is intolerable.
190
usrbinkat.io @usrbinkat.io · 15/05/2026
Okay brief chitchat about slop machine clanker usage. I do use llm quite a bit for technical / verifiable tasks. Something like comparing implementation architectures is nice to rubber duck with RAG on the upstream source code and an LLM. I find it reasonably accurate. What are y'all doing?
GitHub Gist by usrbinkat containing README.md for "Public IP Address Mapping in Multi-Rack Kubernetes Clusters," showing the 18-section table of contents covering five implementation tiers ranked by Kubernetes/Cilium BGP idiomaticity for resolving Public IP → Instance → Node → ToR Switch mappings on bare-metal clusters.

https://gist.github.com/usrbinkat/6bb637b08212b4d9090cea011e276d3c#decision-matrix
160
usrbinkat.io @usrbinkat.io · 05/05/2026
What if <chat_app>, but cli, tui, gui, webui, serverless... ...via declarative config driven policy governance community chat ops?
3213
usrbinkat.io @usrbinkat.io · 12/04/2026
- Clone an entire github.com/<project>/** - Alt-tab survives compositor reconfig - Closed windows gone from switcher - HiDPI per-output scale tracking - gix+git2: pure Rust, no git CLI - sesame clone auto-discovers workspace - workspace update and init workflows - Grouped workspace
A screenshot shows terminal output from the Open Sesame CLI demonstrating batch repository cloning functionality. The dark-themed terminal displays tabs for bash and ssh with username usrbinkat-optiplexprime and timestamps showing 10:51. The command sesame clone --project https://github.com/ScopeCreep-zip executes from the open-sesame main branch on host mithril.

The output reports Found 11 repositories in github.com/ScopeCreep-zip followed by a vertical list of repository names with existence status. Each repository shows in muted gray text with "(exists)" annotation: website, Rack, kalilix, ossiodata, dollhaus.zip, open-sesame, Symbiont, SpiritStream, ReKindle, and Springtale. The summary line reports 0 cloned, 10 skipped (exist), 0 failed, indicating all discovered repositories were already present locally and no new clones were necessary. The prompt returns to usrbinkat@mithril open-sesame main ready for the next command.

This demonstrates Open Sesame's project-level batch cloning capability that discovers all repositories under a GitHub organization or user URL and clones them in a single operation. The exists detection prevents duplicate cloning by checking local workspace before initiating downloads, enabling efficient workspace synchronization across machines or after system reinstalls. This workflow automation eliminates manual navigation to individual repository pages and repetitive git clone commands, replacing dozens of manual operations with a single command that handles organization-wide repository management intelligently.
140
usrbinkat.io @usrbinkat.io · 12/04/2026
Today was an OpenSesame day. Still playing with the workspace layout and features but it's starting to be extremely useful even for developing itself. Next we'll launch the nix flake devshells seamlessly, no nix skills required.
A screenshot shows terminal output from the Open Sesame CLI running workspace management commands in a dark-themed terminal. The top bar displays bash and ssh tabs with username usrbinkat-optiplexprime and timestamp 08:31. The command sesame workspace list followed by sesame workspace status show their outputs in monospace font with syntax highlighting.

The workspace list displays a hierarchical tree of 20 git repositories organized under 12 organizations across 3 servers. Each server section begins with a URL, followed by indented organizations and repositories. Repositories show name, branch, abbreviated commit hash, and status in green for clean or yellow for dirty. Workspaces are tagged with workspace annotations. Projects include infrastructure and k9 under git.braincraft.io/braincraft, gitoxide under GitoxideLabs, nix and nixpkgs under NixOS, comfy-table under Nukesor, and Springtale, open-sesame marked dirty, rekindle, and spiritstream under github.com/ScopeCreep-zip. Additional repositories span catppuccin, nix-community, siderolabs, wiedymi, qemu-project, and virtio-fs. Summary reads 3 servers, 12 orgs, 20 repos.

The workspace status shows open-sesame repository details: path /workspace/usrbinkat/github.com/ScopeCreep-zip/open-sesame, remote URL, branch main, commit 5662337 with message chore release 1.6.19 skip ci, tracking status origin/main up to date, dirty working tree, no profile, and namespace ScopeCreep-zip with workspace.git indicator.

This demonstrates Open Sesame's workspace management functionality beyond window switching: comprehensive git repository oversight across multiple hosting platforms, providing at-a-glance status of entire project ecosystems with branch tracking and uncommitted change detection for developers managing dozens of repositories simultaneously.
170
usrbinkat.io @usrbinkat.io · 30/03/2026
Spent Saturday at the river and Sunday on ScopeCreep.zip and BrainCraft.io. Then this AM my $dayjob M4 MBP had no signs of life. I thought it was bricked. Then the goddess of geek squad @radicalkjax.com touched it and now everything is better (ofc).
selfie of usrbinkat touching grass (aka outside by the river)
2190
usrbinkat.io @usrbinkat.io · 22/03/2026
I remember Pre-MAGA Tea Party fears over DHS inland invasion via the "100 miles inland rule" exploiting airports for full reach into the interior. I remember Tea Party fear of big government cattle car human detention conspiracies. The Tea Party is the DHS invasion and mass detention force now.
DHS AND ICE ARE FEDERAL TERRORIST ORGANIZATIONS
2151
usrbinkat.io @usrbinkat.io · 18/03/2026
It's a boot scoot boogie 🥾🛴🕺
160
usrbinkat.io @usrbinkat.io · 03/03/2026
I've been on the tmux hype train for over a decade. It's one of the most necessary tools for cloud development. Keep on preaching 🙏 I just got deeper into tmux in tmux for nested local + remote tmux and the quality of life improvement just from consolidating my terminal window count was 🤌
111
usrbinkat.io @usrbinkat.io · 25/02/2026
ETCd was written by an intern at CoreOS. It's amazing how far it's carried us on a planetary scale. 🌍🌎🌏 pierrezemb.fr/posts/notes-...
220
usrbinkat.io @usrbinkat.io · 22/02/2026
2160
usrbinkat.io @usrbinkat.io · 16/02/2026
No man has read this manual.
020
usrbinkat.io @usrbinkat.io · 16/02/2026
The single threaded bottleneck situation improved after RTFM.
110
usrbinkat.io @usrbinkat.io · 11/02/2026
Y'all, the platform is the prompt.
030
usrbinkat.io @usrbinkat.io · 11/02/2026
The future (of cloud) is (browser) management. I'll be speaking at SCALE 23x in LA March 5-8, and for the first time it's my presentation as a technologist instead of a corporate presentation for shareholder value. I really can't wait to see you there! www.socallinuxexpo.org/scale/23x
4305
usrbinkat.io @usrbinkat.io · 10/02/2026
`~$ ssh mithril "docker save busybox" | docker load` I do not like it, but it works.
0171
usrbinkat.io @usrbinkat.io · 08/02/2026
Streaming today streaming today! First step: install the lobsta!!! @freecodecamp.bsky.social how did you slop so fast?
270
usrbinkat.io @usrbinkat.io · 08/02/2026
Considering live streaming some dev time today. Who want's to see what OpenClaw does in Ops? Konductor is 98% there, and fun in the browser now so what if? I'm expecting the worst but I've never tried the lobster bots so anything could happen.🦞🤖
3200