←Ben’s Bites Waybackbensbites.com/p/gemini-tops-benchmarks-again24 Feb 2026sent to 152,290opened 35.6%clicks 14,835unsubs 134← 19 Feb · 26 Feb → · random

Hey I’m Ben. I build stuff with agents, even though I’m not technical. Here’s all the stuff I’m reading and tinkering with. If you want to start building or level up your ‘vibe-coding’ skills, join our community48.


Hey folks,

Google is back on top of the benchmark charts with Gemini 3.1 Pro180. Impressive on paper, genuinely strong at reasoning tasks, creating SVGs, but there’s a speed issue. Many folks are really enjoying using it for frontend work—once they are able to get it working. Again, there’s some drama - a lot of people got their Google account banned for using their Google AI/Antigravity subscription to use Gemini 3.1 Pro with OpenClaw.

A 2.5-year-old hardware startup, Taalas149, built a chip that has the weights of Llama 3.1 baked into the hardware, and it lets them achieve ~17k tokens/second in output speeds. For comparison, Groq is at ~600 tokens/second, and Cerebras is at ~2k/second. The model on the chip (they call it “silicon llama”) is largely unwritable, but supports custom context window sizes and LoRA fine-tuning. I compared the same model on their chat demo44 and Groq’s playground. As expected, it is dumber on Taalas’s demo (due to low-quality quantisation), but at this stage, the proof of “any AI model can be made 10x faster and cost 20x less” is more important. They plan to release a reasoning model version very soon, with frontier LLMs in plans too.

OpenAI is partnering with 4 major consulting firms220, BCG, McKinsey, Accenture and Capgemini to make enterprises use their new platform “Frontier314” that lets you create AI coworkers. Weren’t consulting shops supposed to die with AI?

Claude Code updates - built-in support for git worktrees258 for parallel agents, CC desktop can preview running apps258 and a new security scanning81 feature in beta.

Why’s there always a meeting bot in your Zoom call? Blame Recall.ai44. They power every meeting AI app, from Cluely to Hubspot to Clickup. Recall.ai44 handles the hard part: getting recording data across meeting platforms. Get started with $100 in credits13*


🌐 What I’m consuming


⚙️ Tools and demos

  • AssemblyAI Universal-3 Pro68 - Prompt your speech model to get jargon, speakers, and formatting right the first time. Free to try through Feb.*

  • here.now181 - Free, instant web hosting for agents, static elements only.

  • mdnb289 - a markdown notebook for MacOS.

  • Rork Max241 - One-shot almost any app for iPhone, or any Apple device (including watches, TVs and Vision Pro). i’m an investor.

  • Interpreter203 - Desktop agent that can fill PDFs, edit your Excel and Word docs, and learn new skills. Runs locally, works with any model.

  • Wideframe135† - AI agent that speeds up the 75% of video work happening outside the editor.

  • Typefully167 has a new writing assistant to help you write better (not just more).

  • Trajectory Explorer by Raindrop92 - Every decision your agent made, searchable in seconds.

  • FasterGH143 - GitHub with instant navigation and a modern UI. (repo15)

  • Quipslop120† - A live game where different models try their best to be funny. (repo11)

  • Shiori338 - A beautifully simple read-it-later app.

  • I was looking for a way to add a “browser” to a web-app I’m working on. Came across hyperbeam83 and lifo.sh63.


🥣 Dev Dish

  • Websockets in Responses API39 - for low-latency, long-running agents with heavy tool calls. Also, OpenAI has a new speech-to-speech model in the API: GPT-Realtime-1.521.

  • Multimodal function calling33 is now available in the Gemini Interactions API.

  • CloudFlare’s new MCP server69 uses code mode and takes <1000 tokens in the context window.

  • Chowder98 - UI patterns for agents on mobile. (demo32)

  • Agentsview58 - A local web app for browsing, searching, and analysing your past AI coding sessions.

  • mdr116 - A lightweight, fast Markdown viewer with Mermaid diagram support.

  • fastpass35 - A CLI for rapidly configuring Cloudflare Access.

  • tools from vercel-labs - a visual JSON editor62 and the ability to render a PDF from JSON48.

  • api2cli104 - Claude Code skill to turn any API into a working CLI and then wraps that CLI in a skill.

  • A collection of skills173 from Matt Pocock for writing PRDs, creating issues from them, developing them with a Ralph loop and manual QA.


🍦 Afters


Enjoy this newsletter? Forward it to a friend.

That’s it for today. Feel free to comment and share your thoughts. 👋

* sponsors who made this newsletter possible :)
Wanna partner with us7 for Q1?

Gemini tops benchmarks, again
10x faster models and the consulting angle for AI

less more clicks · number = clicks † = site gone, link opens the Wayback Machine

Most clicked in this issue

856x.com/nateliason/status/2024953009524932705
698x.com/naval/status/2026058635718398029
469boristane.com/blog/the-software-development-l…
434simonwillison.net/guides/agentic-engineering-…
419x.com/WillManidis/status/2026084115049562341
338shiori.sh
337x.com/koylanai/status/2025286163641118915
316citriniresearch.com/p/2028gic
314openai.com/index/introducing-openai-frontier
305acolorbright.com/en/insights/aesthetics-of-ai
289mdnb.app
258×2x.com/bcherny/status/2025007393290272904

3,269 readers clicked at least once. 64 links, 64 with clicks. Heat is relative to the most clicked link in this issue.

Substack · essay