Everything in AI is a hillclimb - algorithms, memory utilization, runtimes, as well as things like validation, software quality, and even user experience. It really makes AI-driven development a lot expensive than initially seemed.
It's never a weekend vibe-coded project.
DSM Pro Engineer retweeted
"The last thing I did is the report on Gaza naming what is happening there as genocide. One of the main reactions to that report [was] paying tribute to me for my courage.. why would you need courage to speak the truth?"
Navi Pillay upon winning the 2026 Nobel peace prize
@lydiahallie There's something seriously wrong with Fable 5.1 usage. It burns up usage like I've never seen. Likely some kind of caching issue as it takes out a chunk when first restoring a session, more so than previously. Can barely go 30 min without hitting 20x session limits.
FIFA REALLY, REALLY wants Argentina to win this World Cup. like, literally every game they rob their opponents to give to Argentina. Thieves.
DSM Pro Engineer retweeted
It took a little longer than expected, but we have created a website for people to view the footage collected from Gaza in one place. You no longer have to download the entire archives to see them.
It includes:
64,537 videos
17,905 photos
Ability to download individual videos
Searchable index
Exhaustive sources list (300+ journalists)
Geolocation data
Livemap with minute to minute updates
Victim list
It can be accessed here: ArchiveGenocide.com
Please share & quote tweet to help this post break out of the twitter algorithm prison.
We will keep adding the rest of the archives to the site, be patient- it is difficult work. Continue to seed the torrents provided, as that is the best way to ensure the footage remains stored in decentalized way.
God bless all those who sacrificed their lives to get this footage out, and everyone invovled in collecting/archiving it.
Join our telegram:
t.me/+p_Ufon9FBOY0Y2Q0
Follow our backup accounts:
@ZionismExposedx & @IsraelExposedAr
DSM Pro Engineer retweeted
UN commission of inquiry says Israel committing genocide in Gaza by deliberately targeting children bbc.in/44m7eF8
GLM-5.2 still crap compared to Opus 4.8. I would say it's a B-/C+ grade. This is debugging on a 50m line codebase. Still, it's doing a LOT better than previous GLM models. It's usable, meaning it can converge on a solution with some instruction effort, but ultimately uninspiring.
if you used Fable 5 to code then you're doing something wrong. Coding is easy, any model cn do that. Instead, use Fable to write your PRDs that the coding agents can then follow. Have Fable design your products architecture.\, not write the code.
Lots of @claudeai 529 Overloaded API Errors Claude Code errors in the middle of Fable Ultracode workflows, losing a ton of very expensive token work when the workflows have to restart.
A good multi-harness pattern is to get Claude Code to manage Antigravity agents by just having Opus write instruction files for simpler tasks for Antigravity to finish using Gemini 3.5 Flash. Gemini 3.5 Flash is better than Haiku, and is really fast. Leave hard stuff for Opus
New Claude Code ultra code mode is great. I have it running several thousands agents, which exceeds the 1,000 agent limit so it has to run it in batches. Let's see if I can take the day off tmrw..
@_mohansolo @_anshulr You guys need to update the @antigravity system prompts so that it doesn't write a script all the time to edit code. Have it directly edit file instead. If it writes scripts, it causes more bugs elsewhere and never converges. Claude Code gets it right.
I tried @antigravity again last night, editing a Python module tree to fix a bunch of bugs pointed out by a linter, and it never converged. Just kept on writing more scripts to edit the code, all night, causing more bugs. This is with Gemini 3.1 Pro (high)