Peter Dedene
@dedene.bsky.social
AI Explorer · Building apps in Ruby, Elixir & NodeJS · Digital Artisan, Musician & Entrepreneur · Partner at Zenjoy
The entire timeline is scrambling to buy new Mac Minis for their Clawdbots.
Meanwhile, this old brick in my closet just realized its retirement is over.
"I thought I was done?"
Not today, buddy. You're a server now.
Meanwhile, this old brick in my closet just realized its retirement is over.
"I thought I was done?"
Not today, buddy. You're a server now.
January 26, 2026 at 8:50 PM
The entire timeline is scrambling to buy new Mac Minis for their Clawdbots.
Meanwhile, this old brick in my closet just realized its retirement is over.
"I thought I was done?"
Not today, buddy. You're a server now.
Meanwhile, this old brick in my closet just realized its retirement is over.
"I thought I was done?"
Not today, buddy. You're a server now.
IPv6 PXE boot on our Proxmox cluster has haunted me for 4 years.
I let Claude Code have a go.
It wasn't magic. It was brutal. But... the thing just doesn't get tired. It grinded through and kept researching and trying.
Finally, it booted.
I’m never touching that config again.
I let Claude Code have a go.
It wasn't magic. It was brutal. But... the thing just doesn't get tired. It grinded through and kept researching and trying.
Finally, it booted.
I’m never touching that config again.
January 22, 2026 at 9:21 PM
IPv6 PXE boot on our Proxmox cluster has haunted me for 4 years.
I let Claude Code have a go.
It wasn't magic. It was brutal. But... the thing just doesn't get tired. It grinded through and kept researching and trying.
Finally, it booted.
I’m never touching that config again.
I let Claude Code have a go.
It wasn't magic. It was brutal. But... the thing just doesn't get tired. It grinded through and kept researching and trying.
Finally, it booted.
I’m never touching that config again.
Soprano TTS is absurdly efficient.
> 80M parameters.
> Under 1GB memory.
> 15ms latency on GPU.
And running entirely on-device.
The creator dropped the training code a couple of days ago, so you can finally fine-tune your own voices locally.
Local interfaces are about to get very chatty.
> 80M parameters.
> Under 1GB memory.
> 15ms latency on GPU.
And running entirely on-device.
The creator dropped the training code a couple of days ago, so you can finally fine-tune your own voices locally.
Local interfaces are about to get very chatty.
January 19, 2026 at 7:56 PM
Soprano TTS is absurdly efficient.
> 80M parameters.
> Under 1GB memory.
> 15ms latency on GPU.
And running entirely on-device.
The creator dropped the training code a couple of days ago, so you can finally fine-tune your own voices locally.
Local interfaces are about to get very chatty.
> 80M parameters.
> Under 1GB memory.
> 15ms latency on GPU.
And running entirely on-device.
The creator dropped the training code a couple of days ago, so you can finally fine-tune your own voices locally.
Local interfaces are about to get very chatty.
This IBM manual from '79 hits harder than most modern AI papers.
January 14, 2026 at 5:49 PM
This IBM manual from '79 hits harder than most modern AI papers.
Third day in a row.
At this point, it feels less like an "opt-in" and more like a war of attrition.
If Anthropic's strategy is to annoy us until we select "ON" and surrender our code, just to make the modal go away, we need a new word for "help improve."
At this point, it feels less like an "opt-in" and more like a war of attrition.
If Anthropic's strategy is to annoy us until we select "ON" and surrender our code, just to make the modal go away, we need a new word for "help improve."
January 11, 2026 at 12:17 PM
Third day in a row.
At this point, it feels less like an "opt-in" and more like a war of attrition.
If Anthropic's strategy is to annoy us until we select "ON" and surrender our code, just to make the modal go away, we need a new word for "help improve."
At this point, it feels less like an "opt-in" and more like a war of attrition.
If Anthropic's strategy is to annoy us until we select "ON" and surrender our code, just to make the modal go away, we need a new word for "help improve."
If you want to see where AI UX is really going, watch what hackers do in the terminal.
David Siegel just gave Claude Code its own display to spawn interactive terminal interfaces for emails, calendars, or markdown documents.
Fast and text-first. Very cool!
github.com/dvdsgl/clau...
David Siegel just gave Claude Code its own display to spawn interactive terminal interfaces for emails, calendars, or markdown documents.
Fast and text-first. Very cool!
github.com/dvdsgl/clau...
January 10, 2026 at 11:12 PM
If you want to see where AI UX is really going, watch what hackers do in the terminal.
David Siegel just gave Claude Code its own display to spawn interactive terminal interfaces for emails, calendars, or markdown documents.
Fast and text-first. Very cool!
github.com/dvdsgl/clau...
David Siegel just gave Claude Code its own display to spawn interactive terminal interfaces for emails, calendars, or markdown documents.
Fast and text-first. Very cool!
github.com/dvdsgl/clau...
January 10, 2026 at 10:32 PM
Asking ChatGPT for facts is about as reliable as asking a fortune teller for lottery numbers.
It looks confident, the vibe is right, but it's mostly just guessing.
It looks confident, the vibe is right, but it's mostly just guessing.
January 10, 2026 at 8:38 PM
Asking ChatGPT for facts is about as reliable as asking a fortune teller for lottery numbers.
It looks confident, the vibe is right, but it's mostly just guessing.
It looks confident, the vibe is right, but it's mostly just guessing.
Just found out Context7 quietly fixed context bloat.
Their new MCP does server-side reranking and returns about 65% fewer tokens per call, with slightly better answers.
Less noise, more signal.
upstash.com/blog/new-co...
Their new MCP does server-side reranking and returns about 65% fewer tokens per call, with slightly better answers.
Less noise, more signal.
upstash.com/blog/new-co...
January 10, 2026 at 6:59 PM
Just found out Context7 quietly fixed context bloat.
Their new MCP does server-side reranking and returns about 65% fewer tokens per call, with slightly better answers.
Less noise, more signal.
upstash.com/blog/new-co...
Their new MCP does server-side reranking and returns about 65% fewer tokens per call, with slightly better answers.
Less noise, more signal.
upstash.com/blog/new-co...
You can't just bolt an agent onto a legacy codebase and hope for the best.
Well, you can. But it won't work.
*Agent-native* is a fundamental architectural shift.
This guide explains exactly how to build for it.
👉 every.to/guides/agen...
Well, you can. But it won't work.
*Agent-native* is a fundamental architectural shift.
This guide explains exactly how to build for it.
👉 every.to/guides/agen...
January 9, 2026 at 3:57 PM
You can't just bolt an agent onto a legacy codebase and hope for the best.
Well, you can. But it won't work.
*Agent-native* is a fundamental architectural shift.
This guide explains exactly how to build for it.
👉 every.to/guides/agen...
Well, you can. But it won't work.
*Agent-native* is a fundamental architectural shift.
This guide explains exactly how to build for it.
👉 every.to/guides/agen...
finally a luxury gift set that matches my brute force workflow
January 8, 2026 at 7:53 PM
finally a luxury gift set that matches my brute force workflow
Taking cc-mirror for a spin today.
The real star of the show? Multi-agent orchestration.
Instead of babysitting one agent, you have a "Conductor" managing a squad of workers (all Claude Codes) via a shared queue.
It feels like cheating to have this much leverage.
The real star of the show? Multi-agent orchestration.
Instead of babysitting one agent, you have a "Conductor" managing a squad of workers (all Claude Codes) via a shared queue.
It feels like cheating to have this much leverage.
January 6, 2026 at 7:05 PM
Taking cc-mirror for a spin today.
The real star of the show? Multi-agent orchestration.
Instead of babysitting one agent, you have a "Conductor" managing a squad of workers (all Claude Codes) via a shared queue.
It feels like cheating to have this much leverage.
The real star of the show? Multi-agent orchestration.
Instead of babysitting one agent, you have a "Conductor" managing a squad of workers (all Claude Codes) via a shared queue.
It feels like cheating to have this much leverage.
RIP MS Office.
RIP Microsoft 365.
Now it's "Microsoft 365 Copilot".
We're one rebrand away from Excel being a prompt box that says "trust me bro, the math checks out."
RIP Microsoft 365.
Now it's "Microsoft 365 Copilot".
We're one rebrand away from Excel being a prompt box that says "trust me bro, the math checks out."
January 5, 2026 at 10:05 PM
RIP MS Office.
RIP Microsoft 365.
Now it's "Microsoft 365 Copilot".
We're one rebrand away from Excel being a prompt box that says "trust me bro, the math checks out."
RIP Microsoft 365.
Now it's "Microsoft 365 Copilot".
We're one rebrand away from Excel being a prompt box that says "trust me bro, the math checks out."
POV: You're 30 seconds into the year 2000 and your Mac still works.
26 years ago this felt like dodging the apocalypse. Now we panic when GitHub goes down for 3 minutes.
The dread is timeless.
26 years ago this felt like dodging the apocalypse. Now we panic when GitHub goes down for 3 minutes.
The dread is timeless.
January 4, 2026 at 5:33 PM
POV: You're 30 seconds into the year 2000 and your Mac still works.
26 years ago this felt like dodging the apocalypse. Now we panic when GitHub goes down for 3 minutes.
The dread is timeless.
26 years ago this felt like dodging the apocalypse. Now we panic when GitHub goes down for 3 minutes.
The dread is timeless.
"Honey, I'm home."
First work day of 2026 and already deep in the same old patterns.
First work day of 2026 and already deep in the same old patterns.
January 2, 2026 at 10:49 AM
"Honey, I'm home."
First work day of 2026 and already deep in the same old patterns.
First work day of 2026 and already deep in the same old patterns.
This is the year I finally switch to Vim.
January 1, 2026 at 9:17 PM
This is the year I finally switch to Vim.
My 2026 resolutions tier list is done.
You can probably guess which tier has the most items.
You can probably guess which tier has the most items.
January 1, 2026 at 1:08 PM
My 2026 resolutions tier list is done.
You can probably guess which tier has the most items.
You can probably guess which tier has the most items.
My plans for New Year's Eve?
Closing all my browser tabs. Every single one.
I'm terrified.
Closing all my browser tabs. Every single one.
I'm terrified.
December 31, 2025 at 9:16 PM
My plans for New Year's Eve?
Closing all my browser tabs. Every single one.
I'm terrified.
Closing all my browser tabs. Every single one.
I'm terrified.
For all TUI enjoyers, if you haven't tried gh-dash.dev, you're missing something. It's a simple gh cli extension but it feels like giving GitHub a proper tiling window manager.
Simple but so good.
Simple but so good.
December 31, 2025 at 2:59 PM
For all TUI enjoyers, if you haven't tried gh-dash.dev, you're missing something. It's a simple gh cli extension but it feels like giving GitHub a proper tiling window manager.
Simple but so good.
Simple but so good.
They said AI would take our jobs.
Nobody mentioned it would take our bloodlines too.
Nobody mentioned it would take our bloodlines too.
December 28, 2025 at 10:41 AM
They said AI would take our jobs.
Nobody mentioned it would take our bloodlines too.
Nobody mentioned it would take our bloodlines too.
My Codex Wrapped 2025 ✨
You can clearly see when Opus 4.5 came around. But I got and still get a lot of value from my Codex subscription!
You can clearly see when Opus 4.5 came around. But I got and still get a lot of value from my Codex subscription!
December 27, 2025 at 2:24 PM
My Codex Wrapped 2025 ✨
You can clearly see when Opus 4.5 came around. But I got and still get a lot of value from my Codex subscription!
You can clearly see when Opus 4.5 came around. But I got and still get a lot of value from my Codex subscription!
Ugh, I hate when Opus 4.5 does this.
It couldn't find the root cause, so it decides on its own to rewrite the test with stubs to force a pass. It didn't fix the bug. It just silenced the alarm.
Green lights, zero value. 🫠
It couldn't find the root cause, so it decides on its own to rewrite the test with stubs to force a pass. It didn't fix the bug. It just silenced the alarm.
Green lights, zero value. 🫠
December 22, 2025 at 10:33 PM
Ugh, I hate when Opus 4.5 does this.
It couldn't find the root cause, so it decides on its own to rewrite the test with stubs to force a pass. It didn't fix the bug. It just silenced the alarm.
Green lights, zero value. 🫠
It couldn't find the root cause, so it decides on its own to rewrite the test with stubs to force a pass. It didn't fix the bug. It just silenced the alarm.
Green lights, zero value. 🫠
I never a single person using the Metaverse.
December 21, 2025 at 8:02 PM
I never a single person using the Metaverse.
GitStory is basically Spotify Wrapped for your impostor syndrome.
It turns your commit history into a movie and you suddenly realize:
“Oh, I did ship. A lot.” 🎬
👉 gitstory.sitestash.org/
It turns your commit history into a movie and you suddenly realize:
“Oh, I did ship. A lot.” 🎬
👉 gitstory.sitestash.org/
December 19, 2025 at 4:02 PM
GitStory is basically Spotify Wrapped for your impostor syndrome.
It turns your commit history into a movie and you suddenly realize:
“Oh, I did ship. A lot.” 🎬
👉 gitstory.sitestash.org/
It turns your commit history into a movie and you suddenly realize:
“Oh, I did ship. A lot.” 🎬
👉 gitstory.sitestash.org/
The Gemini 3 update is live. Flash, Thinking, Pro.
Time to feed it some messy prompts and see where it sweats. 🤩
Time to feed it some messy prompts and see where it sweats. 🤩
December 17, 2025 at 4:17 PM
The Gemini 3 update is live. Flash, Thinking, Pro.
Time to feed it some messy prompts and see where it sweats. 🤩
Time to feed it some messy prompts and see where it sweats. 🤩