Dmitriy Anderson

Agentic AI Program Management Director at Adeo

[shawn @shawn_pana]

Introducing: Phone Harness MCP 🔌

Any agent that supports MCP can now control a real iPhone or Android in the cloud:

> works in Claude, ChatGPT, Cursor, Codex, and more as a custom connector
> nothing to install, just paste a URL and sign in
> sees the screen, taps, types, and swipes like you do

Available in phone-harness 0.3.3.

https://t.co/88u1gSgvtS

[Éros Brousson @erosbrousson]

GENIUS 🤣🤣🤣

He created a script to counter the new data tax the French government wants to adopt:

- He downloads a PDF file available on the Socialist Party's website on an infinite loop, which racks up €30 in taxes in 4 hours.

- In theory, if 1,000 people did this, they could drain their €16 million budget in 5 months 🤣🤣🤣

The law doesn't exempt associations or political parties, so honestly, part of me kind of wants it to pass 🤣🤣🤡🤡🤡

[Addy Osmani @addyosmani]

Claude Code tip: if a task turns out harder than you expected, type ultrathink while Claude is working. It raises the effort for the rest of that turn. In the terminal on the latest release.

[Louis Anslow @LouisAnslow]

AI. IS. UNPOPULAR. BECAUSE. IT. BREAKS. CREATIVE. PROFESSIONALS. ILLUSION. THEY. ARE. SPECIAL.

[Nicholas Fabiano, MD @NTFabiano]

The more you fake it, the less you fake it.

Repetition rewires the brain for competence and identity.

[Sahil Jada @BySahilJada]

My new favourite fonts to unslop designs.

Save this for later 👀

[nader dabit @dabit3]

Turns out human slop was 1,000x worse than AI slop.

[Ben Davis @davis7]

Highly recommend connecting your t3 code machines to grok bot. I’ve been doing more and more work from my phone because of it

Also works great with cursor cloud agents (esp the projects setup). With the cursor setup it’s a two way thing where grok bot and cursor can message back and forth.

T3 code isn’t quite as seemless yet but it’ll get there…

[Prasenjit Sarkar @stretchcloud]

Anthropic published the code review setup the Claude Code team runs internally this week. Merged PRs more than doubled, reverts dropped over 40%, and PRs with real review comments went from 16% to 54%. The part that stood out to me wasn't the REVIEW.md file or the checklist. It was step three: several agents hunt for bugs in parallel, a separate group of agents votes on each finding, and it only gets posted if enough of them agree it's real. Style nits get dropped before a human ever sees them.

That's the exact mechanic I built into Campfire's permission layer, just applied one step earlier, before the change even merges instead of after. Run Claude Code, Codex, Goose, Aider, and OpenHands side by side in one browser tab, each in its own git worktree, and when an agent wants to do something risky, a majority of the group has to approve it, any single deny blocks it, and there's a 30 second timer so nobody's agent sits there waiting on a human who stepped away. Session replay shows exactly which agent said what and why. You can also race the same task across every backend in isolated worktrees and see which one actually produced the better diff instead of guessing from vibes.

Single-agent review, one model marking its own homework, is what Anthropic's own numbers show was the weak setup. Multiple agents voting, with disagreement required before anything gets trusted, is what actually moved the metrics. That's not just a code-review insight. It's an agent-governance insight, and it applies just as much to the agent writing the change as the ones reviewing it.

https://t.co/WwOLZbMwgh
https://t.co/9Ue6NxVPl3

[Daniel San @dani_avila7]

I'm slowly enabling auto-merge across my repos

Claude Code agents already handle PRs, CI failures, security scans, and AI code reviews (Cubic + Greptile + Claude Code review + NVIDIA Skillspector)

Once all checks pass, PRs can merge automatically

IMPORTANT: Sensitive changes still require manual approval

One less bottleneck when maintaining open source projects with AI agents

[Ahmad @TheAhmadOsman]

Last mile delivery in software is gonna become the moat

[gabriel @gabriel1]

anyone who is great at deeply explaining things now can make 1h long videos explaining any deeply complex topics in a highly intuitive way

like someone could produce a 1h video about how markets work and how things get cheap and affordable, show it at school, and end communism once and for all

[Theo - t3.gg @theo]

There are hundreds of “ai dev influencers” who share their thoughts, tools, skills, etc.

There’s maybe 10 who actually know what they’re talking about

[Armin Ronacher ⇌ @mitsuhiko]

I'm convinced that at this point, you could have an Opus 5.5 agent re-implement all of CPython probably within two weeks without any human input, against just the tests, and probably with full compatibility with including ABI, no GIL, a jit and probably beat it on all measurable metrics. It's completely crazy.

[log4j @log4j]

i’m hearing reports it’s over, but i’m feeling we’re incredibly back?

[toli @tolimarchuk]

The new 5-inch Macbook Pro has touchscreen and 5G.

[sankalp @dejavucoder]

opus 5.5

[Andrej Karpathy @karpathy]

Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5:
https://t.co/JIhXN7OSh1

- Excellent recs without any logins (state is kept in local browser session)
- Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads
- Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version
- HackerTube Longs ✊

A bit more on the process:
I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing.

Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.

[Chris Tate @ctatedev]

What shipped at Vercel Labs in October

95 releases
730 issues closed
3,500+ PRs merged

New: agent-desktop
→ Computer use for any coding agent
→ macOS, Windows + Linux

New: tsc-ts
→ The TypeScript 7 compiler, written in TypeScript
→ Compiled to native with scriptc

agent-browser
→ Safer MCP
→ 42 browser tools for eve agents

skills
→ Sync skills across a monorepo
→ Sync reads optional + peer deps
→ Faster updates

json-render
→ Streaming fixes
→ Form + PDF fixes
→ json-render-shadcn skill

portless
→ No more sudo
→ portless mcp

deepsec
→ Kimi K3.1 Pro + DeepSeek V4.1 Flash benchmarks

native
→ native test for TS apps
→ Reusable markup components
→ Virtual lists

scriptc
→ Fully native, no JS engine
→ WebAssembly compiles to native
→ Profile-guided optimization

just-bash
→ mktemp + yes
→ Closer to real bash + jq

fx
→ Launches 23× faster
→ 41% fewer tokens
→ Agents that survive redeploys

wterm
→ Faster color rendering
→ Emoji + flags in one cell
→ Vim matches your page theme

vgpu
→ Build-time WGSL shaders
→ Turbopack support
→ Depth + MSAA on canvas

lat․md
→ lat ui to browse your knowledge graph
→ Graph view + static export

emulate
→ npx emulate mcp
→ AWS Cognito
→ Slack interactivity

ai-cli
→ ai mcp
→ ai decide can judge images

mcp-handler
→ MCP Events

run
→ CPU time limits for agent code

[tobi lutke @tobi]

When I first told @MParakhin that we were planning to move to a monorepo he said wisely:

Moving to a Monorepo is an idea that every engineer absolutely hates, and that once its there, no engineer would ever want to work at a company that doesn't use one. We will definitely do it.