Augustinas Malinauskas

I asked Claude Code to build 100 games with one prompt and this is what it came up with. Try games: classic-games-arcade.ml.ink Check out for token usage, duration, what worked well and what didn't 👇 @claudeai @AnthropicAI

@hthieblot @mldotink

Can this new X algo find incredible builders/founders? If that’s you tell me what you are building.

@1752vc @mldotink

Founders 👋 Looking to connect with more people working on: AI Tech Growth hacking SaaS Building in public Drop what you’re working on below👇

@NVIDIAGeForce #007FirstLightRTX

Recruits, your first prize is here... A custom GeForce RTX 5080 Founders Edition + PC copy of the game. Comment #007FirstLightRTX to win 👇

@IterIntellectus @MayorofLondon can you please act

what the fuck is happening in the UK?! an 18yo kid was stabbed four times by a fellow british* man "Vikrum Digwa" who was openly carrying an 8 inch blade in public (which is illegal in the UK, but under the new racial laws only one ethnic group is allowed to do) chased down when he tried to escape, stabbed again, but since the gentleman claimed henry did one of the worst crimes one can commit in the west, doing a racism, the police put the bleeding kid in handcuffs and let him die on the pavement what the fuck?!

@engineers_feed Progress bar for any goal

What hasn't been invented, but needs to be ASAP?

@AniC_dev Can you share objective reasoning of why?

honestly he should resign

@chamath Nice work, but RAG should be in action layer rather than intelligence, since it’s often used with tool calls

@adahstwt Research on other platforms, buy on aws or cloudflare - they don’t seem scammy. Squarespace is the worst product in existence, very sad that googledomains sold

hey devs, where do you usually buy your domains? 1. GoDaddy 2. Namecheap 3. Cloudflare 4. Porkbun 5. Dynadot 6. Squarespace 7. Hostinger 8. Other

@dabit3 @DevinAI Building cloud run by agents @mldotink

I'm giving away five $200 Max plans for @DevinAI If you're using tools like Claude Code, Codex, or Cursor and haven't tried @DevinAI, comment below with what you're building to be eligible. ⚡️

@AishwaryaDevv Or just ask agent to do it with @mldotink

For people who keep asking how to deploy a simple static HTML personal blog: 1. Buy a .io domain for $70/year 2. Provision 3 AWS EC2 instances 3. Setup a multi-region Kubernetes cluster 4. Write 4,000 lines of Terraform config 5. Deploy a high-availability Kafka event stream 6. Rewrite the backend API in Rust 7. Containerize everything with Docker 8. Set up Grafana and Prometheus for observability 9. Implement an enterprise-grade CI/CD pipeline 10. Configure a distributed Redis caching layer 11. Setup PagerDuty alerts for 99.999% uptime 12. Realize the site only gets 1 visitor a month 13. Forget how to center a div in CSS 14. Abandon the project completely

@ManOfFocus_ @grok fact check

Elon Musk was right when he said, "Go Get your fucking confidence back. Without it, you will fail miserably no matter how skilled or smart you are."

@opencode Does Go store conversations?

OpenCode x DeepSeek V4 Flash - free for a limited time DeepSeek V4 Flash is currently our most popular model in Go Give it a try if you haven’t already

Newton proved this first. Imagine a planet made of shells/layers like an onion. Newton proved that if you’re inside a shell you feel no gravity, because different parts of the shell pull you and overall effect cancels out. Hence as you approach center of the earth the shells behind do not contribute to the net force pulling you. This is consequence of inverse squared law and applies to electric force too.

Why gravity decreases with depth? #Science

@crypt0lake You need power both underground and overground. In space there is a flux of photons for solar panels - hence power is free. The main question how much power will be needed and how much more efficient hardware and software gets

why is no one making underground data centers

@eladgil Everyone from NY to London is using claude code and codex now. Good tools propagate quickly now. Perhaps you want to give specific example, because it’s not true

People at major AI labs (using internal models) 3-4 months ahead of startup silicon valley engineers SV founders/eng 3-6 months ahead of NY NY founders/eng 6-12 months ahead of rest of world Most people have no idea how fast AI shifting as 1-2 years behind SOTA "The future is here, just not equally distributed" - Robert Heinlein

@chadiek @siddak_b @thdxr The goal is to build a unified API that can be used with any provider. Cache configuration is different across providers. For example Anthropic requires explicit opt-in with either 5min or 60min ttl. As one builds such unified API have to take differences into account.

how does kv cache configuration fit here though? prompt caching is tied to a specific provider and model, so if you're switching between providers (or even models within the same provider) you're getting a cold cache every time regardless. the abstraction layer can't really help with that. can you please explain

we're working on a library to abstract over all the llm providers there's very few teams that have dealt with the quirks between providers at the scale we have it's written in effect but will also have a vanilla api progress is in the opencode repo under packages/llm

KV cache configured differently, reasoning effort has different labels and levels, streaming sse comments, different server side tools, strict ignored in tool definitions, usage details are different, temperature limits are different, anthropic completions api doesn’t support caching - have to use official, anthropic will concatenate all system messages, anthropic will not return reasoning in completions API. This is not even exhaustive list. For example OpenRouter abstracts both completion and responses APIs to something that looks like completions.

@thdxr Honestly, I'll admit I haven't built anything too complex with these abstraction libraries/APIs. But OpenRouter seems to manage this decently well already. What's it missing? Bit confused what some of these "quirks between providers" are

@CR1337 The moment one cannot bypass ads, the people have lost

YouTube (Premium) is now blocking Mullvad IPs: "To continue, turn off your VPN / proxy. This will alow YouTube to locate the best content."

@MatthewBerman @trq212 People using models daily been taking advantage of this for 6 months already

Imagine writing an article on X and changing an entire industry instantly. @trq212 has powerful ideas.

@hiarun02 Comparison between ChatGPT, Grok, Gemini

How to trick Opus 4.7 into thinking.