zed

Cross chain swapping should be effortless, we are close! @PineappleDex

Pure vibe coding is a Ponzi scheme. Eventually, the technical debt comes due, and if you don't understand the foundations, you can't pay it off.

@RoundtableSpace Cheap for a reason

OPEN SOURCE MODEL ALTERNATIVES FOR CLOSED MODELS: * OPUS 4.6 - GLM 5 / MINIMAX M2.5 * GPT 5.2 HIGH - KIMI K2.5 THINKING * GPT-5.2 - QWEN3.5-397B * GEMINI 3.1 PRO - DEEPSEEK V3.2 SPECIALE * NANO BANANA - QWEN IMAGE EDIT

@bridgemindai You can get a refund on the ultra sub due to this

Gemini 3.1 Pro is struggling. Throughput on Google Vertex dropped to 50 TPS. It was hitting 92 TPS at launch. The model that impressed me last week is now slow and inconsistent. Simple tasks are taking longer. Performance is degraded. Google is having a rough rollout. This keeps happening with their releases. I'm still bullish on Gemini 3.1 Pro. The intelligence is there. But they need to fix the infrastructure. Hoping Google gets this together soon.

@googledevs @geminicli Maybe fix the gemini plans first, ultra unusable errors non stop, agent manager is even worse

Announcing the launch of Hooks in @GeminiCLI 🚀🪝 Take full control and customize the agentic loop to your specific needs. Hooks allow you to add context, validate actions, enforce policies, loop the agent to continue to iterate, and more. Read the blog → goo.gle/4rDEDFm

@mattpocockuk I heard you mention tmux, have you tried zellij, it's awesome 👌

All the investments I've made in my ralph loops and skills are really paying off now. I am producing more high-quality code than ever and there are still so many low-hanging fruit to pick. Super-exciting times to be a dev.

@forloopcodes He also said he pushes straight to main 🤢

openclaw devs are so dumb why dont they use openclaw to fix all the issues, merge all prs, and remove all security flaws in one shot after all, the creator of this slop already agrees he never reads what the llm does

@RoundtableSpace Are you trolling

Grok 4.20 or Claude Opus 4.6, which are you picking?

@pankajkumar_dev Can confirm this, tested heavily for 2 days

Antigravity Problems: - Eats a lot of RAM and feels very heavy. - Quotas finish too fast now. Need API key option, pay-as-you-go, or a plan between Pro and Ultra. - Agent terminated due to error happens too frequently. - Models don't perform as well compared to Claude Code or Codex, especially with long context. - If history gets large, it sometimes doesn't even open. Have to wait or terminate it. - UI/UX needs improvement no clear view of remaining context or token usage, and no compact context feature. - Browser feature is very slow. Navigation and screenshots take too much time. - Overall performance has become slow, especially with Gemini models. Please google team fix these issues and refine it.

I come back to this video once in a while...

Literally every crypto influencer has pivoted to AI. Absolute shiny object syndrome. Most contagious disease known to man.

@AniC_dev don't knock it till ya try it youtube.com/watch?v=1ZfO14…

@ze_dev over ssh? not sure I could do that on windows sounds wild

introducing ADHD coding spawn unlimited agents in parallel each is working in isolation from its own computer in the cloud producing changes we can truly interact with, feel, debug via desktop streaming now we just switch desktop, criticize, switch desktop, criticize, merge

@AniC_dev yes, use a desk/web app setup

@ze_dev can you test desktop apps this way no

@BogdanVadadav @bridgemindai It's like ordering a Nokia phone from Alibaba packaged as an iPhone

@ze_dev @bridgemindai yeah, i tried to make a documentation page for my software with glm-5 and deepseek 3.2, after 5 hours of ai slop i decided to switch back to opus 4.6 and it fixed it in 30mins...

Qwen 3.5 Plus just dropped with a 1 million token context window. $0.40 input. $2.40 output. Per million tokens. For context: Claude Opus 4.6: $5/$25 GPT 5.3 Codex: $1.75/$14 GLM 5: $1/$3.20 Qwen 3.5 Plus: $0.40/$2.40 Hybrid architecture combining linear attention with sparse mixture-of-experts. Native vision-language capabilities built in. Alibaba says it matches state-of-the-art on task evaluations while being 10x cheaper than the frontier labs. The open source community has been waiting for this one. Qwen models consistently punch above their weight and this is their most ambitious release yet. 1M context at this price point changes the math on a lot of workflows. Time to benchmark it.

people who say "ai is gonna replace developers" are failed tech bros who: a) never built anything b) failed to learn coding c) could never get a job in tech d) have no idea what software development means lashing out for their own shortcomings

The “I built a SaaS in a weekend” crowd needs to stop. Show me one that doesn’t look like it was designed by someone who just discovered Tailwind. Same font. Same layout. Same rounded boxes. Same colour scheme. Same landing page with the hero, three features, and a pricing table. It’s a homogenization machine. And that’s just the surface. Try going off the happy path. Try entering something unexpected. Try using it like a real person who doesn’t know how it’s “supposed” to work. Watch it break. No error handling. No edge cases considered. No security. Just vibes and a deployed Vercel link. Cool, you shipped something in 48 hours. But you didn’t build a product. You built a demo that happens to accept credit cards (and I wouldn’t put my card in there). It’s the equivalent of a band recording a demo in a garage and thinking they got it made. Building fast is great, but building fast and calling it done is how we end up with an internet full of shit that all looks the same and barely works.

AI is not ready.

@darshal_ It's only good for coding

STOP OVERPAYING $200 for CLAUDE Kimi K2.5 is now the #1 most used model for OpenClaw & THE MOST USED MODEL OVERALL on OpenRouter! it has Claude Opus4.6 performance, but 8x CHEAPER!!! Here's everything you should know: