Ed Lepedus's Avatar

Ed Lepedus

@elepedus.bsky.social

AI Engineer, Elixir enthusiast

588 Followers  |  827 Following  |  74 Posts  |  Joined: 05.11.2024  |  2.0475

Latest posts by elepedus.bsky.social on Bluesky

Post image

So... it's official!

I'm building an identity provider.

wildcard.id is the start of my resistance. If you're a builder, entrepreneur or user looking for a different vision for the future of technology, join me.
#buildinpublic #elixir

30.10.2025 12:45 โ€” ๐Ÿ‘ 5    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Next time youโ€™re building something with Claude Code, add this snippet to your prompt:

โ€œLetโ€™s implement this using strict outside-in TDD, whereby we write the simplest possible failing test, followed by the smallest amount of implementation code that will satisfy it, and iterate from thereโ€

01.08.2025 12:57 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Phoenix Playground

I made another thing! KinoReverseProxy is an #Elixir package that lets you give your #Livebook apps their own domain name! ๐Ÿš€

See it in action at www.speedrun.dev!

hex.pm/packages/kin...

02.03.2025 20:32 โ€” ๐Ÿ‘ 12    ๐Ÿ” 1    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Aaaaand.. now Iโ€™ve fixed the stupid bug I left in my first release! #facepalm

17.02.2025 17:29 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Preview
kino_phoenix_live_view Allows for Phoenix LiveView integration with Livbook.

Just published a little #elixir library to make it easier to use #Phoenix #Liveview in #Livebook!

Try `KinoPhoenixLiveview.new(path: "/proxy/apps/{slug}")`

hex.pm/packages/kin...

16.02.2025 20:16 โ€” ๐Ÿ‘ 6    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Post image

Letโ€™s go!! ๐Ÿš€

Iโ€™m so pumped to be running a microhack for the #Montreal #Elixir guild tonight!

Thanks Cleaver Barnes for inviting me!

guild.host/events/micro...

29.01.2025 18:10 โ€” ๐Ÿ‘ 3    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Still works, yet to need a charge.

23.01.2025 17:46 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Median Digit

Yep. I made mediandigit.com ๐Ÿคฆ

30.12.2024 19:57 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Fair

23.12.2024 13:44 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Counterpoint: commit to doing 1 minute of exercise every day, and to increasing that by 1 minute each week.

Went from 0 to 51 mins/day this year, and clocked up over 11,000 minutes of exercise.

23.12.2024 11:45 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

Imagine teaching programming using #Elixir and #Livebook.

First lesson is building a chat room, and the rest of the course is delivered through it ๐Ÿš€

20.12.2024 21:18 โ€” ๐Ÿ‘ 5    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

๐Ÿ“Œ

13.12.2024 11:50 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

This is beautiful!

13.12.2024 11:50 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

๐Ÿค” Imagine #Kubernetes, but instead of pods, the basic unit of deployment is a #Livebook app...

#elixir #devops

13.12.2024 11:05 โ€” ๐Ÿ‘ 3    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Post image

Why I love #Elixir...

13.12.2024 11:04 โ€” ๐Ÿ‘ 4    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Unfortunately, the financial model has encouraged a lot of lazy click-bait trash.

At one point, only 1 in 12-15 medium articles I opened were genuinely useful, and more often than not, I'd run out of free reads by the time I found something good.

12.12.2024 18:21 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

๐Ÿ˜ญ

12.12.2024 18:17 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

๐Ÿ“Œ

12.12.2024 18:16 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

I thought Iโ€™d be fine with 2TB, but dammit, the models, they just keep getting MOAR BIGGERRER!

Now Iโ€™m eyeing up the 4TB and 8TB upgrades and questioning my sanity! ๐Ÿคฃ

12.12.2024 15:12 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

Yeah, I hear you!

It was especially frustrating when I first got my Mac and enthusiastically enabled all the SDKs, then a while later started running out of disk space and would have to juggle things around to make room every time there was a point release ๐Ÿคฆ

12.12.2024 15:01 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

Haha! You dodged a bullet there! Imagine downloading 25GB on the venue WiFi minutes before a talk! ๐Ÿ˜ฌ

12.12.2024 14:24 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

FWIW, I made a concerted effort this year to stop clicking Medium links.

Lots of great people still write there, but I find the Medium gating so obnoxious, Iโ€™m actively boycotting it.

12.12.2024 14:05 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

Whatโ€™s wrong?

12.12.2024 13:58 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

Also, with real people rather than astroturfing bots!

12.12.2024 09:56 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
A sleep tracking dashboard showing data for 6 hours and 17 minutes of sleep. The โ€˜Sleep Stagesโ€™ chart illustrates time spent in Awake (3 hours), REM (1 hour 21 minutes), Light sleep (3 hours 40 minutes), and Deep sleep (1 hour 15 minutes) stages across the night. A โ€˜Slow Wave Intensityโ€™ chart shows fluctuation in intensity with 143 restoration points. A โ€˜Sleep Positionโ€™ chart indicates the user spent 37% of their sleep on their left side, with periodic changes in position. The timeline spans from 12 AM to 8 AM.

A sleep tracking dashboard showing data for 6 hours and 17 minutes of sleep. The โ€˜Sleep Stagesโ€™ chart illustrates time spent in Awake (3 hours), REM (1 hour 21 minutes), Light sleep (3 hours 40 minutes), and Deep sleep (1 hour 15 minutes) stages across the night. A โ€˜Slow Wave Intensityโ€™ chart shows fluctuation in intensity with 143 restoration points. A โ€˜Sleep Positionโ€™ chart indicates the user spent 37% of their sleep on their left side, with periodic changes in position. The timeline spans from 12 AM to 8 AM.

First thoughts on the Muse S band: Holy sh*t! *This* is #sleep tracking! Oura take note!

12.12.2024 09:52 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Preview
Music to Focus Better - Brain.fm Use Brain.fm's functional music to help you focus relax and sleep better, using cutting edge neuroscience to get results.

Subscription fatigue has killed my love of paid software. I remember as a student scraping together ยฃ300 for the Adobe suite, ยฃ20-200 for OmniPlan, Sketch, DevonThink etc and using them for years.

Now I pay my JetBrains sub, Brain.fm and thatโ€™s about it. Most sub/ppu apps I instantly dismiss.

11.12.2024 22:22 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

I really like this -- thanks for sharing! :)

11.12.2024 17:54 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
This image illustrates a comparison of AI reasoning between two scenarios, labeled โ€œThenโ€ and โ€œNow.โ€

On the left, titled โ€œThen,โ€ OpenAIโ€™s best hosted model at the time is asked, โ€œMikeโ€™s mum had 4 kids; 3 of them are Luis, Drake, and Matilda. What is the name of the 4th kid?โ€ The model initially responds, โ€œIt is not possible to determine the name of the fourth child without more information.โ€ Even when prompted further with โ€œThe answer is in the question,โ€ the model fails to infer the context and insists that the answer cannot be determined, showing a lack of reasoning ability for this type of riddle.

On the right, titled โ€œNow,โ€ a small, open-source model (running locally for free on a laptop) is asked a similar question: โ€œErinโ€™s mum had 4 kids; 3 of them are Bobby, David, and Alan. What is the name of the 4th child?โ€ This model correctly deduces and explains, โ€œThe answer is โ€˜Erinโ€™!โ€ It identifies that Erin is one of the four children mentioned in the question, demonstrating an improved ability to infer context and reason through riddles.

The contrast highlights significant progress in AI reasoning, with a smaller, freely accessible model outperforming a previously leading proprietary one in this specific scenario.

This image illustrates a comparison of AI reasoning between two scenarios, labeled โ€œThenโ€ and โ€œNow.โ€ On the left, titled โ€œThen,โ€ OpenAIโ€™s best hosted model at the time is asked, โ€œMikeโ€™s mum had 4 kids; 3 of them are Luis, Drake, and Matilda. What is the name of the 4th kid?โ€ The model initially responds, โ€œIt is not possible to determine the name of the fourth child without more information.โ€ Even when prompted further with โ€œThe answer is in the question,โ€ the model fails to infer the context and insists that the answer cannot be determined, showing a lack of reasoning ability for this type of riddle. On the right, titled โ€œNow,โ€ a small, open-source model (running locally for free on a laptop) is asked a similar question: โ€œErinโ€™s mum had 4 kids; 3 of them are Bobby, David, and Alan. What is the name of the 4th child?โ€ This model correctly deduces and explains, โ€œThe answer is โ€˜Erinโ€™!โ€ It identifies that Erin is one of the four children mentioned in the question, demonstrating an improved ability to infer context and reason through riddles. The contrast highlights significant progress in AI reasoning, with a smaller, freely accessible model outperforming a previously leading proprietary one in this specific scenario.

Sometimes it's easy to forget just how far we've come!
#AI #LLM #reasoning

11.12.2024 17:50 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Its weird how, in the land of freedom, health, education and justice are most likely to bankrupt you!

11.12.2024 08:56 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Nice ๐Ÿ‘Œ Looking forward to hearing about your journey and discoveries!

07.12.2024 22:42 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

@elepedus is following 18 prominent accounts