bruncs

software engineer • building @ScrunchAI • prev. @coinbase @itau

[Julia Turc @juliarturc]

We need an official name for not-vibe-coding. My top candidates so far:

· boomer coding
· chewgy coding
· trad coding
· Coding with capital C

Suggestions are welcome.

[Scrunch @ScrunchAI]

Tired of scrolling past 60+ mentions of “chicken” before finding the actual recipe? 🥣

That was the SEO era—writing for search engines, not for humans.

The new AI-first era rewards clarity, accuracy & usefulness.

We made a fun little video to show why the #chickensoup era of search optimization is over.

[Scrunch @ScrunchAI]

We're excited to announce our Series A led by @DecibelVC with participation from @Mayfield and Homebrew. This raise will accelerate our ambitious mission to build the first AI customer experience platform. Read more about what we're building: https://t.co/EHWu5FkWvj

[Kane 謝凱堯 @kane]

I wonder if these PMs know and hate each other in real life

[chris andrew @chriswandrew]

Your website doesn't need to go away, but 90% of its human traffic will.

[lisatomic @lisatomic5]

are you kidding me

[bruncs @bruncs]

@Jason @ScrunchAI 🥇

[kitze @thekitze]

gemini 2.5 in cursor after saying "Sure, I'll work on that"

[Harj Taggar @harjtaggar]

Autists had a great run, the AI future belongs to ADHD

[Italo @peppipets]

Nunca na história desse país teve um governo tão pró banqueiro quanto esse do Lula

Usar FGTS do trabalhador pra pagar empréstimo de 3% de juros ao mês é muito amor aos bancos

Refute-me

[Jared Palmer @jaredpalmer]

Anthropic made the right move.

MCP's can now be fully stateless, resumable, and implemented with just plain HTTP with this RFC.

(link in reply)

[Austen Allred @Austen]

What is it about really really smart people that make them completely oblivious to how incredible AI is?

[Bytebytego @bytebytego]

Top 4 Forms of Authentication Mechanisms

1. SSH Keys:
Cryptographic keys are used to access remote systems and servers securely

2. OAuth Tokens:
Tokens that provide limited access to user data on third-party applications

3. SSL Certificates:
Digital certificates ensure secure and encrypted communication between servers and clients

4. Credentials:
User authentication information is used to verify and grant access to various systems and services

Over to you: How do you manage those security keys? Is it a good idea to put them in a GitHub repository?

--
Subscribe to our weekly newsletter to get a Free System Design PDF (158 pages): https://t.co/eVEdOFSYPY

[Rust Language @rustlang]

Rustup 1.28.0 has been released with many improvements. ✨

For those using rustup in CI: Note that rustup will no longer automatically install the toolchain from `rust-toolchain.toml`. Make sure to run `rustup toolchain install` if necessary.

Blog post: https://t.co/9mnOoWLpci

[Theo - t3.gg @theo]

The Material Theme has just been removed from GitHub and VS Code due to shipping malware

[CZ 🔶 BNB @cz_binance]

Waiting for the new headline: #Bitcoin "CRASHES" from $1,001,000 to $985,000.

Save the tweet.

[Michael Saylor @saylor]

₿itcoin on $ale

[Alex Cheema @alexocheema]

Market close: $NVDA: -16.91% | $AAPL: +3.21%

Why is DeepSeek great for Apple?

Here's a breakdown of the chips that can run DeepSeek V3 and R1 on the market now:
NVIDIA H100: 80GB @ 3TB/s, $25,000, $312.50 per GB
AMD MI300X: 192GB @ 5.3TB/s, $20,000, $104.17 per GB
Apple M2 Ultra: 192GB @ 800GB/s, $5,000, $26.04(!!) per GB

Apple's M2 Ultra (released in June 2023) is 4x more cost efficient per unit of memory than AMD MI300X and 12x more cost efficient than NVIDIA H100!

Why is this relevant to DeepSeek?
DeepSeek V3/R1 are MoE models with 671B total parameters, but only 37B are active each time a token is generated. We don't know exactly which 37B will be active when we generate a token, so they all need to be ready in high-speed GPU memory.

We can't use normal system RAM because it's too slow to load the 37B active parameters (we'd get <1 tok/sec). On the other hand GPUs have fast memory but GPU memory is expensive. Apple Silicon, however, uses Unified Memory and UltraFusion to fuse dies - a tradeoff that favors a large amount of medium-fast memory at a cheaper cost.

Unified memory shares a single pool of memory between the CPU and GPU rather than having separate memory for each. There's no need to have separate memory and copy data between the CPU and GPU.

UltraFusion is Apple's proprietary interconnect technology for connecting two dies with a super high speed, low latency connection (2.5TB/s). Apple's M2 Ultra is literally two Apple M2 Max dies fused together with UltraFusion. This is what enables Apple to achieve such a high amount of memory (192GB) and memory-bandwidth (800GB/s).

Apple M4 Ultra is rumored to use the same UltraFusion technology to fuse together two M4 Max dies. This would give the M4 Ultra 256GB(!!) of unified memory @ 1146GB/s. Two of these could run DeepSeek V3/R1 (4-bit) at 57 tok/sec.

All of this and Apple has managed to package this in a small form-factor for consumers with great power efficiency and great open-source (uncharacteristic of Apple!) software. MLX (h/t @awnihannun) has made it possible to leverage Apple Silicon for ML workloads and @exolabs has made it possible to cluster together multiple Apple Silicon devices to run large models, demonstrating DeepSeek R1 (671B) running on 7 M4 Mac Minis.

It's unclear who will build the best AI models, but it seems likely that AI will run on American hardware, on Apple Silicon.

[bruncs @bruncs]

@tlakomy 🇧🇷 Brazil: 50 Mbps ~ USD 19

[Henning Dieterichs @hediet_dev]

https://t.co/x23SFl8Xnu in VS Code. Don't leave your editor for creating or updating your diagrams.

https://t.co/OPSXAm3TgN