Adrian Pilarczyk
@peelar.bsky.social
Solution Architect at Saleor Commerce
peelar.dev
peelar.dev
No way I'd have built these a year ago. Now I'm writing this with Claude Pen, while walking my dog, and the image was made in dopeshot.
That's personal software for you.
BTW. If you think dopeshot could help your screenshot game, DM me 💬
That's personal software for you.
BTW. If you think dopeshot could help your screenshot game, DM me 💬
December 10, 2025 at 2:58 PM
No way I'd have built these a year ago. Now I'm writing this with Claude Pen, while walking my dog, and the image was made in dopeshot.
That's personal software for you.
BTW. If you think dopeshot could help your screenshot game, DM me 💬
That's personal software for you.
BTW. If you think dopeshot could help your screenshot game, DM me 💬
2️⃣ Claude Pen
I live in my terminal, so I moved my writing there too.
It takes my voice-dictated ramblings—typos and all—runs them through an LLM pipeline I control. Cleans up the mess, reviews what's strong and weak. It's learned my style, so I trust it.
Try it: github.com/peelar/clau...
I live in my terminal, so I moved my writing there too.
It takes my voice-dictated ramblings—typos and all—runs them through an LLM pipeline I control. Cleans up the mess, reviews what's strong and weak. It's learned my style, so I trust it.
Try it: github.com/peelar/clau...
GitHub - peelar/claude-pen: Command-based pen pal inspired by Claude Code agentic workflows
Command-based pen pal inspired by Claude Code agentic workflows - peelar/claude-pen
github.com
December 10, 2025 at 2:58 PM
2️⃣ Claude Pen
I live in my terminal, so I moved my writing there too.
It takes my voice-dictated ramblings—typos and all—runs them through an LLM pipeline I control. Cleans up the mess, reviews what's strong and weak. It's learned my style, so I trust it.
Try it: github.com/peelar/clau...
I live in my terminal, so I moved my writing there too.
It takes my voice-dictated ramblings—typos and all—runs them through an LLM pipeline I control. Cleans up the mess, reviews what's strong and weak. It's learned my style, so I trust it.
Try it: github.com/peelar/clau...
1️⃣ dopeshot
Cool products deserve cool screenshots. But making them shouldn't require a Figma degree.
Paste your assets, get good colors and layout automatically, export something that looks decent. That's it.
Try it: dopeshot.vercel.app
Cool products deserve cool screenshots. But making them shouldn't require a Figma degree.
Paste your assets, get good colors and layout automatically, export something that looks decent. That's it.
Try it: dopeshot.vercel.app
dopeshot
Your product is dope, your screenshots should be too. Drop a screenshot, get a Twitter-ready graphic in 10 seconds.
dopeshot.vercel.app
December 10, 2025 at 2:58 PM
1️⃣ dopeshot
Cool products deserve cool screenshots. But making them shouldn't require a Figma degree.
Paste your assets, get good colors and layout automatically, export something that looks decent. That's it.
Try it: dopeshot.vercel.app
Cool products deserve cool screenshots. But making them shouldn't require a Figma degree.
Paste your assets, get good colors and layout automatically, export something that looks decent. That's it.
Try it: dopeshot.vercel.app
2. ElevenReader (elevenreader.io)
turns articles into audiobooks. voice quality is top notch and you can even select the preferred voice
I just had Michael Caine explain to me how legacy codebases can be indexed by LLM-based solutions. it's a good life
turns articles into audiobooks. voice quality is top notch and you can even select the preferred voice
I just had Michael Caine explain to me how legacy codebases can be indexed by LLM-based solutions. it's a good life
Free Text Read Aloud App | ElevenReader
Read text aloud with ElevenReader app. Listen to free audiobooks and read aloud PDFs, eBooks, and Kindle with the highest-quality TTS voice AI.
elevenreader.io
December 2, 2025 at 10:55 AM
2. ElevenReader (elevenreader.io)
turns articles into audiobooks. voice quality is top notch and you can even select the preferred voice
I just had Michael Caine explain to me how legacy codebases can be indexed by LLM-based solutions. it's a good life
turns articles into audiobooks. voice quality is top notch and you can even select the preferred voice
I just had Michael Caine explain to me how legacy codebases can be indexed by LLM-based solutions. it's a good life
1. Wisper Flow (wisprflow.ai)
voice-to-text that actually works. way better than iOS dictation.
my main use? dictating prompts to Claude/Cursor/ChatGPT. you end up giving 10x more context when you just talk vs type.
voice-to-text that actually works. way better than iOS dictation.
my main use? dictating prompts to Claude/Cursor/ChatGPT. you end up giving 10x more context when you just talk vs type.
Wispr Flow | Effortless Voice Dictation
Flow makes writing quick and clear with seamless voice dictation. It is the fastest, smartest way to type with your voice.
wisprflow.ai
December 2, 2025 at 10:55 AM
1. Wisper Flow (wisprflow.ai)
voice-to-text that actually works. way better than iOS dictation.
my main use? dictating prompts to Claude/Cursor/ChatGPT. you end up giving 10x more context when you just talk vs type.
voice-to-text that actually works. way better than iOS dictation.
my main use? dictating prompts to Claude/Cursor/ChatGPT. you end up giving 10x more context when you just talk vs type.
It can strip you of any delusion that people at the helm of companies that target children must be more empathetic, with a sense of mission and care about this vulnerable type of user.
Nope, it's all about the number going up.
Nope, it's all about the number going up.
December 1, 2025 at 10:14 AM
It can strip you of any delusion that people at the helm of companies that target children must be more empathetic, with a sense of mission and care about this vulnerable type of user.
Nope, it's all about the number going up.
Nope, it's all about the number going up.
It is quite apparent that the company cares more about virtue-signaling to stakeholders rather than actually speaking to their users or their parents.
If you have children around you who spend time playing games online, I recommend watching this interview, no matter how uncomfortable it is.
If you have children around you who spend time playing games online, I recommend watching this interview, no matter how uncomfortable it is.
December 1, 2025 at 10:14 AM
It is quite apparent that the company cares more about virtue-signaling to stakeholders rather than actually speaking to their users or their parents.
If you have children around you who spend time playing games online, I recommend watching this interview, no matter how uncomfortable it is.
If you have children around you who spend time playing games online, I recommend watching this interview, no matter how uncomfortable it is.
The moment that shocked me the most was when the CEO claimed multiple "famous parents" reached out saying their children "would be dead" if it wasn't for friends they found on Roblox. Like, what? This is your response to allegations you're not doing enough to protect children?
December 1, 2025 at 10:14 AM
The moment that shocked me the most was when the CEO claimed multiple "famous parents" reached out saying their children "would be dead" if it wasn't for friends they found on Roblox. Like, what? This is your response to allegations you're not doing enough to protect children?
www.youtube.com/watch?v=XpI...
The interview was full of posturing, gushing over the scale they operate at, the sophistication of their moderation tools (while being vague when asked for details), and the numbers Roblox is hitting.
The interview was full of posturing, gushing over the scale they operate at, the sophistication of their moderation tools (while being vague when asked for details), and the numbers Roblox is hitting.
We Asked Roblox's C.E.O. About Child Safety. It Got Tense. | EP 163
Today Roblox has more than 150 million daily active users, and functions as one of the primary online gathering places for preteens. But precisely because Ro...
www.youtube.com
December 1, 2025 at 10:14 AM
www.youtube.com/watch?v=XpI...
The interview was full of posturing, gushing over the scale they operate at, the sophistication of their moderation tools (while being vague when asked for details), and the numbers Roblox is hitting.
The interview was full of posturing, gushing over the scale they operate at, the sophistication of their moderation tools (while being vague when asked for details), and the numbers Roblox is hitting.
It doesn't make them automatically guilty, but what matters is that the company is under heavy scrutiny at the moment.
The company's CEO was invited to one of my favorite podcasts, @, to explain their position on child safety. His response and attitude were perplexing, to say the least.
The company's CEO was invited to one of my favorite podcasts, @, to explain their position on child safety. His response and attitude were perplexing, to say the least.
December 1, 2025 at 10:14 AM
It doesn't make them automatically guilty, but what matters is that the company is under heavy scrutiny at the moment.
The company's CEO was invited to one of my favorite podcasts, @, to explain their position on child safety. His response and attitude were perplexing, to say the least.
The company's CEO was invited to one of my favorite podcasts, @, to explain their position on child safety. His response and attitude were perplexing, to say the least.
Can confirm. I love Anthropic models but def not their apps
November 26, 2025 at 2:18 PM
Can confirm. I love Anthropic models but def not their apps
I plan to set up observability to capture what the LLM handles, which lets me iteratively improve the deterministic layer.
Still, it's cool to be able to set it up that way from the start instead of either spending weeks building more robust pipeline or just having to deal with false negatives.
Still, it's cool to be able to set it up that way from the start instead of either spending weeks building more robust pipeline or just having to deal with false negatives.
November 26, 2025 at 12:03 PM
I plan to set up observability to capture what the LLM handles, which lets me iteratively improve the deterministic layer.
Still, it's cool to be able to set it up that way from the start instead of either spending weeks building more robust pipeline or just having to deal with false negatives.
Still, it's cool to be able to set it up that way from the start instead of either spending weeks building more robust pipeline or just having to deal with false negatives.
The pattern is straight-forward:
1. Run strict, deterministic checks first.
2. If confidence drops or classification fails, hand the input to an LLM that returns structured output.
1. Run strict, deterministic checks first.
2. If confidence drops or classification fails, hand the input to an LLM that returns structured output.
November 26, 2025 at 12:03 PM
The pattern is straight-forward:
1. Run strict, deterministic checks first.
2. If confidence drops or classification fails, hand the input to an LLM that returns structured output.
1. Run strict, deterministic checks first.
2. If confidence drops or classification fails, hand the input to an LLM that returns structured output.
For those of you who've been building with AI for a while, this might seem obvious but it was quite a discovery for me. We’re entering a world where you no longer need to over-engineer brittle heuristics because the cost of handing edge cases to a model is now low enough to be justifiable.
November 26, 2025 at 12:03 PM
For those of you who've been building with AI for a while, this might seem obvious but it was quite a discovery for me. We’re entering a world where you no longer need to over-engineer brittle heuristics because the cost of handing edge cases to a model is now low enough to be justifiable.
Then, it struck me: since the operation is high-stakes and system already has access to an LLM, why not use it as a fallback instead of forcing everything through deterministic logic?
November 26, 2025 at 12:03 PM
Then, it struck me: since the operation is high-stakes and system already has access to an LLM, why not use it as a fallback instead of forcing everything through deterministic logic?
The domain is wide, the patterns are inconsistent, and every time I tried to solve it with regex or rule-based logic, I ended up patching one more edge case on top of the previous one.
November 26, 2025 at 12:03 PM
The domain is wide, the patterns are inconsistent, and every time I tried to solve it with regex or rule-based logic, I ended up patching one more edge case on top of the previous one.
Reymont masterfully captured that feeling of the world shifting beneath you. You're forced to reinterpret your place, terrified of becoming obsolete.
Want to see what history teaches us about navigating technological disruption? Read more 👇
↳ open.substack.com/pub/peelar/...
Want to see what history teaches us about navigating technological disruption? Read more 👇
↳ open.substack.com/pub/peelar/...
November 24, 2025 at 10:16 AM
Reymont masterfully captured that feeling of the world shifting beneath you. You're forced to reinterpret your place, terrified of becoming obsolete.
Want to see what history teaches us about navigating technological disruption? Read more 👇
↳ open.substack.com/pub/peelar/...
Want to see what history teaches us about navigating technological disruption? Read more 👇
↳ open.substack.com/pub/peelar/...
The book follows three entrepreneurs in Łódź, Poland during the Industrial Revolution. In a single lifetime, this city exploded from under 1,000 people to 500,000. How? Thanks to steam-powered textile machinery that made entire trades obsolete in a decade.
November 24, 2025 at 10:16 AM
The book follows three entrepreneurs in Łódź, Poland during the Industrial Revolution. In a single lifetime, this city exploded from under 1,000 people to 500,000. How? Thanks to steam-powered textile machinery that made entire trades obsolete in a decade.
If you are trying to wrap your head around how the social fabric can be warped by a huge technological shift, "The Promised Land" (that's the English title), published originally in 1899, can offer an interesting perspective.
November 24, 2025 at 10:16 AM
If you are trying to wrap your head around how the social fabric can be warped by a huge technological shift, "The Promised Land" (that's the English title), published originally in 1899, can offer an interesting perspective.