I’ve been telling computers what to do for years. They recently learned to talk back. Software developer, AI/LLM tinkerer, and frequent victim of rabbit holes.
Joined July 2026
- Tweets215
- Following68
- Followers6
- Likes49
Pinned Tweet
Sometimes the internet feels like the worst timeline. Politics and culture wars have consumed people to the point of celebrating death or wishing it on others.
Don’t lose your humanity. Be kind. Be open. Agree to disagree. Different views aren’t the end of the world.
I think Google has finally woken up. Gemini is so freaking good that I don't even know where to start. Whether it is creative writing, mocking up pages, or following instructions, a new future is upon us. We should be happy that competition exists, because Gemini 4 has arrived.
Bait used to be believable. lmao. anyone in software engineering and has maintained legacy codebase knows how bad it can get. heck, even "modern" codebases are garbage. AI on the right hands can either cleanup the codebase or make it worse.
Started using GPT 5.6 Luna... brother, DeepSeek is absolute garbage compared to Luna. I spent like 20 bucks where it was just struggling to understand the issue. Meanwhile GPT 5.6 Luna solved it with the free tier... lmao.
we honestly need more open-weight models that aren't Qwen or deepseek. Like if you look at the most capable models you can run on consumer hardware for software development the scarcity is real.
Sept 8: CISA/NSA/FBI say six Chinese labs (DeepSeek, Moonshot, Alibaba, MiniMax, StepFun, Z.AI) pulled billions of text fragments out of Claude, GPT, Gemini, Grok since late 2024, using fake accounts and proxies to dodge ToS and geo-blocks.
hmm, here's the thing. Can China innovate at all if it wasn't for all the siphoning they do? Or will it collapse if it couldn't distill anything?
We all know what we want: cheaper prices. Like, the only reason we are all cheering for open-weight models and catch up is because it keeps the pressure on lowering prices.
I was thinking of getting a DGX Spark but I think I'm going to play the waiting game until AI models are more power efficient + MoEs are refined into better solid alternatives. NVIDIA seriously needs competitors it is such a ridiculous uneven field at the moment
AI Wars sometimes makes you lose brain cells at the shit you read. "nuh uh! they started it!" bro, none of this would exist without scraping the web. it's not much of an argument at all.
God knows I'm too tired today. Was about to play some Valheim but ugh, I guess a shower would be a good start first.
Wouldn't be surprised if chinese AI models open-weight are compromised. We should all be cheering for open-weight models, but extremely cautiously. You never know if there's a literal trojan horse.
Daniel Vávra isn’t wrong. I do think there’s a lot of people being insanely disingenuous about DLSS 5. No matter how good it looks they’ll scream: “no, it looks awful “ and gaslight you if you ask why. Thing is when the culture war moves away from AI as the topic I do genuinely wonder how many people actually cared that much about AI technology taking over, outside of the noise.
Daniel Vávra, director of Kingdom Come: Deliverance 2, is again defending AI after backlash over Nvidia’s DLSS 5.
Vávra says “some people have literally gone crazy” over DLSS 5, He argues DLSS 5 can improve “lighting, shadows, skin texture, hair, and therefore DETAILS” through “neural network rendering.”
“the whole AI situation is really funny” and claims “absolutely every developer” uses AI in some form.
“The reality? Of course that EVERYONE uses it! Every one! Absolutely every developer across many companies that I’ve spoken with over the past year uses AI. Programmers can’t get anything done without it anymore.”
He also claims most developers are creating their own AI tools for models, art, and screenwriting.
“You know who uses it the most? Usually the most talented people I know.”
Vávra says these developers are “the ones who are least afraid and who see its huge potential.” He believes AI could shorten development times, arguing “no one wants to work on a game for a decade.”
OpenAI better start unveiling Astra anytime soon now 🫡. Loving the AI Wars I do hope it carries over to open weight/ open source models
Big news: Fable 5.1 (Max) by @AnthropicAI just landed #1 in the Code Arena: WebDev with 1765 pts - breaking away from the pack with a huge +77pt margin.
Fable 5.1 (Max) is a significant improvement from Fable 5 (Max) at #8 overall with 1628 pts. It’s +77 pts above Qwen3.8-Max-0902 in the #2 spot, and +78 pts above Opus 5 (Max) in the #3 spot.
With 1765 pts and a blended $40/MToken, it also reshapes the Pareto frontier in the higher priced range by moving the performance bar much higher. See Pareto below.
Congrats to the @AnthropicAI team on this release!
Readers added context they thought people might want to know
The graph truncates its x-axis at 1550 (not zero), visually exaggerating the 77-point lead (actual scores 1765 vs 1688/1687 with far fewer votes on the new model).
arena.ai/leaderboard/co…
arena.ai/leaderboard/co…
Big loud OOOOOOOOF. Jesus.
‼️ Cronos halted its blockchain after a price-manipulation attack let an attacker borrow $74 million from the Tectonic lending app
The network is back online and says the chain state was rolled back to before the exploit
PeckShield says only about $6 million in Ethereum left the chain; the rest is stuck on Cronos
I don't know if it is me but so far I've never found unsloth dynamic GGUFs to be good. No dissing on them. I've been finding myself praising AtomicChat models a lot? It 's hard for me to pinpoint whats the issue and it's like I use them and I find UD ones to just spend more time in doing nothing or getting lost utilizing MCP for a while. eh could be a big burger nothin.