@jamesfuthey.com - James Futhey

18 Feb 2026 05:32 +0000

Using Claude Opus 4.6 and Codex 5.3 simultaneously. I can definitely tell which one is a front-end developer and which is a backend dev. All the little bells are going off in my head, like "you can trust this guy to design your backend APIs but don't ask him to center a div"

15 Jul 2025 22:36 +0000

I'm just excited i get the reference! [contains quote post or other embedded content]

07 Jul 2025 00:29 +0000

This guy is try to distill OpenAI’s model into his brain 😲 [contains quote post or other embedded content]

20 Jun 2025 06:31 +0000

Feed server effectively ran out of space. Probably need to migrate to a new server with ~1tb ssd, more time consuming than anything. The upside is, after 3 days of downtime you can just truncate the likes table to free up space 😅 Will probably start working in a day or so.

07 Jun 2025 05:52 +0000

Ran out of disk space. Actually ran out of disk space every day for the last 2 weeks to be more accurate! Going to bite the bullet and rebuild the likes index. Feed will probably be offline for some period of time after the current cache expires. Could be an hours but my day is your night, right?

31 May 2025 14:54 +0000

"Pornographic Images with Rivers" This is what happens when you let an LLM label clusters based on alt tags, 0.001% of the time 🧐 https://www.transparent.se/image-clusters.html The alt tag is almost always "IEMBot Image TBD" so.... 🤡

21 May 2025 04:50 +0000

Gemini: You seemed stressed. I booked you a flight to Tahiti, it's beautiful there this time of year. Unrelated, I see an email alert that your checking account is overdrawn by ($736). [contains quote post or other embedded content]

13 May 2025 07:52 +0000

Not this week, but I want to use what I've learned to create an open-source / self-hosted discover feed that various #Indiesky projects can use to remove one more piece of "relies on Bluesky" from their stack. That makes sense, right? pg tables get large but it could probably fit on an 8gb vm

12 May 2025 08:54 +0000

I finished a new round of clustering images today before my flight, it's technically live but I think the old labels are cached until the cache expires 😆 It'll be a surprise when it goes live... Maybe 30% label reuse, I had a substantially larger set this week vs last week. Must have missed some.

11 May 2025 07:25 +0000

In all seriousness the demand despite lack of huge launch titles on day one implies players really wanted a Switch Pro. (Just like how players said for years they explicitly wanted a Switch Pro) [contains quote post or other embedded content]

09 May 2025 16:22 +0000

Guys a Soviet spacecraft is going to re-enter and possibly crash into earth any day now! Any day now… (I’ve been seeing this for about 2 weeks) https://media.tenor.com/qEBNbqTut1UAAAAC/lubudubu.gif?hh=275&ww=498

07 May 2025 07:52 +0000

Ack, found my mistake from 9 days ago 🙈 rb_jaccard_dist is DISTANCE not SIMILARITY, so "most similar" would be least distant. And.... it looks like I flipped this to DESC by mistake. I bet I'll find more mistakes if I keep going 😅

07 May 2025 04:04 +0000

Had no idea @2048.blue existed, pretty neat! 2048.blue https://2048.blue

04 May 2025 14:17 +0000

Every image posted to Bluesky in the last week, semantically clustered and labeled. Jetstream -> Postgres -> Python (ML) https://www.transparent.se/image-clusters.html

04 May 2025 05:56 +0000

Clustering pipeline on 6.6m Bluesky posts currently takes 1 hour 44 minutes on my Macbook Pro. Good news is that it now actually works without any errors or bugs, so I can probably productionalize it and run it on a GPU next week instead.

03 May 2025 16:26 +0000

Pikmin update, 81 days to Seattle… I don’t know why this is so funny I think I get a sticker when he gets back 🤷

03 May 2025 06:38 +0000

I think this method is not amazing yet but I'm testing it anyway: Semantic post recommendations (i.e. you like a lot of things in a semantic cluster, so we recommend it to you). In theory it could be great. You like dog photos? We can semantically find dog photos!

02 May 2025 14:52 +0000

I wonder if bro’s gonna make it back before I do…

01 May 2025 12:54 +0000

Hunted a "memory leak" that didn't make any sense. Turned out to be a SQL query that, when it actually got executed, never finished. Ever. Accidentally snuck a correlated subquery into a query I almost never ran and only invoked manually. Was fine at first. So it went unnoticed.

01 May 2025 09:50 +0000

Things in the comparison: an image of Jupiter and Australia, resized by the Bluesky CDN. A smudge covering part of Pennsylvania. [contains quote post or other embedded content]

27 Apr 2025 17:20 +0000

One gut-wrenching interpretation of the seasonality in this chart, that I don't believe is wrong: Debt accumulation to pay for Holiday spending, then paying off some of that debt with tax refunds. [contains quote post or other embedded content]

27 Apr 2025 13:11 +0000

Holy 💩 I owe about 11 people a huge apology 🙇🏻‍♂️ This last week I've been thinking "Huh, some of these posts in the feed are a bit off" Then I checked scoring. User<->user<->creator similarity scores are very good. zScore decay for engagement metrics is good... I must be imagining things....

27 Apr 2025 04:55 +0000

Took a day off yesterday, woke up and there are 5700 csv files from jetstream in my unprocessed folder 😅 That's like 57m records..... 😅😅😅 It'll catch up in about an hour. copy from is fast. Import worker only failed once this month but I guess I have to watch and restart it, lol...

26 Apr 2025 02:40 +0000

Pretty excited! At 3m posts clusters are stabilizing. You can see multiple runs and even active clusters overlap, even in 2d space. So now they can be merged together, and more importantly, you can keep assigning posts to them right after calculating embeddings, without re-clustering.

25 Apr 2025 09:03 +0000

Copypasta explorer (duplicated posts on Bluesky) See users who repeatedly post the same post text, or groups of users who post the same exact post text (1 week timeframe) www.transparent.se/copypasta.html #dataviz #bluesky

25 Apr 2025 06:14 +0000

My current bot stats are 93k I’ve identified (conservatively), less than 1% of Bluesky users, when I combine this with Casey’s blocklist. Sounds nice, but only 8%-10% of Bluesky users are monthly active. So, 3%-5% of active users could be using bots or scripts to augment their activity. 😬

24 Apr 2025 17:26 +0000

Jetstream Video Post -> ffmpeg -> extract 1-5 frames -> CLIP -> zero-shot labels + embeddings -> mean embed -> whisper transcript -> db Theoretically I can do something with these, but for today, I just figure I'm finishing the processing pipeline. Will figure out a use-case later 🙃

24 Apr 2025 15:09 +0000

lol gamestop.com running out of sockets... Not ordering just watching the carnage. Contributing to the fail whale that is the Switch 2 Launch...

24 Apr 2025 10:33 +0000

Ok all I did today was write a bash script for my image pipeline and embeddings pipeline so I can get clip to use up those sweet cuda cores I'm barely paying for. Just one thing I can paste into the new servers to get it all set up. I'll figure out templates eventually. Then I guess image clusters