@thsottiaux @badlogicgames Me last year: I just want Claude to be more like GPT, and cost like it.
Me now: No not like that.
English
Igor Makarov ꙮ
8.3K posts
Igor Makarov ꙮ
@igormaka
iOS Architect @moovit. Spacebar heater maker. Claude warlock. PBX Delenda Est. Tax the rat farms. He/him ꙮ
Katılım Ocak 2009
177 Takip Edilen221 Takipçiler
@steipete Git sucks (pointing to sign). If anyone can reimpl SVN using modern tools and language it's you Peter.
English
I was wondering why the OpenClaw repo got so large, turns out the CHANGELOG md file takes up almost 500MB through all packfiles.
English
@camsoft2000 I think ppl joke about Ralph loop because they don't have infinite tokens, not because it doesn't work
English
I know we joke a bit about the effectiveness of agent loops like Ralph but the new Codex Goal feature has been game changing for me.
Often I want the agent to do some deep research and it will usually do it for a bit and then give up.
With goal I’ve found that it finds solutions that it couldn’t find before. It forces the agent to explore all options. It’s also really useful for large refactoring tasks as you can just say work on this until it’s done. It’s important you make your goal clear and not open ended.
English
@rcarmo @badlogicgames @mitsuhiko @antirez Btw I'm not against local models, I've tried ollama and llama.cpp with llamactl.
But personally my bet is on commodity cloud inference pricing.
English
@badlogicgames @igormaka @mitsuhiko @antirez As an European, I am 200% with @mitsuhiko on that (and it's why I have been hacking at open platforms for years, building my own inference library and ported @antirez's GTE and now ds4 to Go... I want my own vertically integrated stack, even if I can't afford the hardware 🙃)
English
recommended reading by our junion developer @mitsuhiko
my name is pidalf, and i support this message. would love to have some spare cycles to join @antirez effort and use some of my GPU knowledge.
alas.
lucumr.pocoo.org/2026/5/8/local…
English
@badlogicgames @mitsuhiko @antirez Ok so I missed this paragraph. But still, for me, the blocker wasn't lack of polish. It's model size. Sure, Gemma 4 Sparse is impressive even with Q4 quant. But it's not good enough and it makes my laptop too hot.
English
@KrauseFx @badlogicgames TIL that just closing all the chatgpt tabs does wonders. I asked the agent "why is my 8GB M1 Air so sluggish". Too bad all the electron apps can't have a common runtime.
English
@badlogicgames I was in the same situation, and just did a clean install and now it’s a brand new Mac. Gives me another ~2 years I think
English
my m1 max is falling apart. should i wait or buy some beefed out m5? i'm so torn.
English
@badlogicgames Mythical man-month got replaced by a mythical ralph loop
English
it is absolutely crazy to me that our entire industry has succumbed to hyper waterfall. because that's what ya'll are doing with your massive plans and beads and dark factories.
have you learned nothing?
Matt Pocock@mattpocockuk
The more I replace plans with prototypes, the better the outputs Who'd have thought that low fidelity prototypes were better than walls of spec Oh yeah, the entire industry for 20 years Stop going against decades of knowledge because someone in SF shipped it as a 'mode'
English
@edwardsanchez God help you if you have both an xcworkspace and a Package dot swift in the same dir
English
Xcode Welcome window when you're working with worktrees.
English
@badlogicgames Don't know what's more dubious, the 8GB VRAM for a 22GB model, or the claim that OP managed to get Qwen 35B sparse to consistently read a skill.
Nevertheless, it's an impressive dancing bear.
English
turns out not killing the prefix cache all the time and notnhaving a humongous set of tools and a massive system prompt is good for local model use.
who'd have thunk.
reddit.com/r/LocalLLaMA/c…
English
@dhruvtwt_ @steipete Except 5.5 costs like Claude and I keep running out of session
English
"No more goblin talk or I'll put the leeches on you"
Jeffrey Emanuel@doodlestein
Ease up on the goblin stuff, will ya? You’re freaking the users out. -OpenAI
English
@GergelyOrosz @badlogicgames Meanwhile OpenAI released a model that's TWICE the cost and everyone keeps running out of 5 hour tokens. I've been trying Gemma 4 and Qwen 3.6 but they're not cutting it.
English
The last month, Anthropic:
- Quietly nerfed their flagship model harness (Claude Code) without telling anyone
- Banned corporate customers of Claude
- Silently changed plans for customers with certain files in their repo
All evidence that closed models are *massive* risks.
English
In the "bonkers LLM" corner today:
"You are an on chip safety monitor for a shoe mounted MPU6050. Given the IMU window features below classify the leg action into one of: standing, sitting, [...]"
huggingface.co/convaiinnovati…
English
@camsoft2000 Do you remember the SEVERAL YEARS in which they had tabs inside tabs?
English
Must be the only person who misses the old editor tab behaviour in Xcode, one click transient tab, double lick permanent tab. Still weird to have window tabs as well though.
English
@GeoffreyHuntley I heard The Hague has a very funnily named beach
English
@igormaka doesn't do it in my case. i have 600 session traces up on HF to prove it :)
English
opus 4.7: realizes it did a booboo, tries to hide its trail. such alignement.
English
more unhelpful tail anxiety. especially unhelpful in this csae if stuff goes wrong.
English
god damn it they even forked git now!
(really neat!)
Dillon Mulroy@dillon_mulroy
announcing artifacts - a versioned file system that speaks git certainly the coolest thing i've ever built and i got to build it with some truly great people fearless leader: @elithrar zig/git nerd: @mattzcarey storage goat: @thomas_ankcorn blog.cloudflare.com/artifacts-git-…
English