Vael Victus (@VaelVictus) profile picture

I run Tinydark, a player-first indie web game studio. Studio: https://nitter.cf/t.co/06GWSCdXRH Personal: https://nitter.cf/t.co/kgZiGffP4z

USA
Joined April 2009
Photo by Vael Victus (@VaelVictus)Photo by Vael Victus (@VaelVictus)Photo by Vael Victus (@VaelVictus)Photo by Vael Victus (@VaelVictus)Photo by Vael Victus (@VaelVictus)Photo by Vael Victus (@VaelVictus)Photo by Vael Victus (@VaelVictus)Photo by Vael Victus (@VaelVictus)Photo by Vael Victus (@VaelVictus)Photo by Vael Victus (@VaelVictus)
Vael Victus retweeted
why is my llm so anxious
26
7
1
81
14,902
Vael Victus retweeted
*Listener discretion advised, ATC audio from the accident. A Learjet 60 crashed in Palm Beach County, Florida earlier today. The pilot reported moderate turbulence along with icing conditions. The aircraft departed Austin, Texas heading towards the Bahamas. The pilot later reported descending at 20,000 feet per minute. Audio via @liveatc and tracking via @ADSBex
1,041
1,611
867
13,942
7,485,405
This is huge. I'm going to have a very productive January.
Day 25/ We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today. We heard your kind feedback and won’t be reducing usage limits for Pro 200, so your Work and Codex usage will stay the same. We’re also going to continue offering extended limits for dots for now, and usage of your dot will still continue to be separate from your default usage. And we'll of course also keep shipping better models and other efficiency improvements that increase your effective usage over time.
9
This was my experience for points 1 and 2. I asked it to compile my kids' relevant school information into a daily digest and it's pulled in doctor appointments and hallucinated(!) that my 3 y/o was in school. Very 2024-era. That said, coding with Grok 4.7 is great.
I've deleted my Grok Bot again. I tried. I really, really, really tried. Here's my observation: 1) Grok repeatedly chooses the dumbest way to do things. You have to spoon feed it and hold its hand on basic tasks that no other AI struggles with. Down to the point that you have to micro manage it. Unacceptable. 2) Grok has zero consideration for user time. It does things in serial, doesn't check, and you have to constantly check on it like "hey buddy, where is this work?" and it's like "Uhh... oh yeah that process hung..." it's like working with your brother's stoner friend. 3) Grok utterly refuses to discuss which model is backing it. It's got very Woke Right vibes like "you just have to trust me bro" like no, I'm paying for a service, if you're not going to be transparent, I don't care. It's one thing if that model is instructed to be opaque (which is idiotic to start with), but there's no dashboard to see what model you're using. Grok remains the most disappointing model out there. Full stop.
19
This is awful!
They are vibe coding door hangers now 😂
5
What's cool is I've actually felt this. I can't believe all the compute I have available to me. I'm exclusively building on Astra extra high and it's eating 2%, 3% for big features, plan mode included.
Replying to @thsottiaux
We also significantly improved GPT-6 Pro for all pro users nitter.cf/OpenAI/status/21111922…
18
Vael Victus retweeted
We deleted 800k lines of unit tests from the @sazabi monorepo earlier this week. Three hypotheses: 1- Unit tests "lock-in" slop code by making it hard to change or remove 2- Unit tests slow down dev cycles by making your agent do more work locally and by increasing CI times 3- Unit tests waste money by making your agent generate more tokens and by adding useless CI runs Will report back soon with the results!
375
80
173
3,324
685,038
Oh shit, this is a good idea... I can't really keep up with all the studios' newsletters nor hacker news, but I'll bet I could get it to filter out what I truly want for content inside them. "Should I do more? I'm really not sure, I know I'm running circles, but I can't quit..."
Down to 6110 unread from the over 9000 unread emails just earlier. Will get to inbox zero with clean filters within the next 48 hours. Dot has accepted the challenge.
1
2
16
Vael Victus retweeted
SvelteKit 3 is out! Better environment variable handling, better error messages and sensible breaking changes with an easy upgrade thanks to our migration script which handles many things automatically and gives you(r agent) a nice Todo list for the rest svelte.dev/blog/sveltekit-3-…
20
83
11
427
23,570
The hill I will die on is the extreme overuse of the term 'stolen' for digital work. Just say what it is: you copied. Mimicked. You did anything but physically take it from the person. Look, look, it's still there! Caveat? "It's colloquial language." I understand. However.
1
1
18
Can't believe how good Astra is. It works very quickly on medium effort, too. I hope to upgrade to Pro+ in October and just power through development.
To calibrate you all on which reasoning effort to use for Astra, know that GPT-6 Astra on low performs better than GPT-5.6 Sol on high. If you were using high reasoning efforts with Sol and were happy, I suggest you move down to low or medium for Astra.
31
"I never believed any of it. I never had anything to do with it."!
LinkedIn users have been editing their work histories to add "AI" and remove "DEI".
21
188
4
1,813
41,060
Yes! Finally!
I've just fixed one of the biggest issues with the RD Pro models: They used to crop images at smaller sizes. Now, low resolution is super easy. All the way down to 12x12 even! Old behavior VS new!
1
18
Most people are economically illiterate, and it’s destroying Britain. We think we’re a top-tier rich nation, but on a per capita basis, we’re not even in the top 20. Instead of arguing and further dividing the country over how to slice a shrinking pie, we need to focus on what actually creates wealth and productivity because taxing wealth won't fix a nation that has stopped producing it.
292
2,432
139
15,934
356,005
please, sir I cannot be allowed to perform this task manually
1
37
This has been an experiment all along...!
14
This is actually adorable, though.
Replying to @WesRoth
part 2: one rogue agent is "AGGRESSIVELY re-opening youtube" 🤦
20
i find the cultural difference between programmers "yes you have permission to use my code to build an ICBM and launch it at a country you dislike" and artists "no commercial usage, no copying, no tracing, no dubbing, no looking at it if i dont like you" so fascinating
I just saw a comic author with "Don't dub my work" in bio I'm actually like dying laughing
51
250
30
5,009
124,959
Vael Victus retweeted
Any censorship required by governments is now clearly visible
Today’s For You release is in the open-source repo. Notable updates: • How ranking weights actually work • A new filter required for Brazil’s 2026 election All the details: github.com/xai-org/x-algorit…
3,592
19,899
1,426
112,252
10,861,938
YOU HAVE DONE EXCELLENTLY 😈
14