Alex Minakov

Investment Director - Red Lions Capital

[Andrej Karpathy @karpathy]

There's a new kind of coding I call "vibe coding", where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It's possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper so I barely even touch the keyboard. I ask for the dumbest things like "decrease the padding on the sidebar by half" because I'm too lazy to find it. I "Accept All" always, I don't read the diffs anymore. When I get error messages I just copy paste them in with no comment, usually that fixes it. The code grows beyond my usual comprehension, I'd have to really read through it for a while. Sometimes the LLMs can't fix a bug so I just work around it or ask for random changes until it goes away. It's not too bad for throwaway weekend projects, but still quite amusing. I'm building a project or webapp, but it's not really coding - I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.

[Andrej Karpathy @karpathy]

It is hard to communicate how much programming has changed due to AI in the last 2 months: not gradually and over time in the "progress as usual" way, but specifically this last December. There are a number of asterisks but imo coding agents basically didn’t work before December and basically work since - the models have significantly higher quality, long-term coherence and tenacity and they can power through large and long tasks, well past enough that it is extremely disruptive to the default programming workflow.

Just to give an example, over the weekend I was building a local video analysis dashboard for the cameras of my home so I wrote: “Here is the local IP and username/password of my DGX Spark. Log in, set up ssh keys, set up vLLM, download and bench Qwen3-VL, set up a server endpoint to inference videos, a basic web ui dashboard, test everything, set it up with systemd, record memory notes for yourself and write up a markdown report for me”. The agent went off for ~30 minutes, ran into multiple issues, researched solutions online, resolved them one by one, wrote the code, tested it, debugged it, set up the services, and came back with the report and it was just done. I didn’t touch anything. All of this could easily have been a weekend project just 3 months ago but today it’s something you kick off and forget about for 30 minutes.

As a result, programming is becoming unrecognizable. You’re not typing computer code into an editor like the way things were since computers were invented, that era is over. You're spinning up AI agents, giving them tasks *in English* and managing and reviewing their work in parallel. The biggest prize is in figuring out how you can keep ascending the layers of abstraction to set up long-running orchestrator Claws with all of the right tools, memory and instructions that productively manage multiple parallel Code instances for you. The leverage achievable via top tier "agentic engineering" feels very high right now.

It’s not perfect, it needs high-level direction, judgement, taste, oversight, iteration and hints and ideas. It works a lot better in some scenarios than others (e.g. especially for tasks that are well-specified and where you can verify/test functionality). The key is to build intuition to decompose the task just right to hand off the parts that work and help out around the edges. But imo, this is nowhere near "business as usual" time in software.

[stepan @cyntro_py]

for three years I've been obsessed with one idea: fully automating my work with AI. and now I'm closer than ever.

the last 1.5 years I used https://t.co/2hef4vH2E3 — a homegrown personal operating system for life and work, built on Cursor. about a month ago I started building version 3 from scratch, now entirely on Claude Code. the results are wild.

my requirements:
- system has access to every digital tool I use: from SMS and email to work CRMs and AI video generators
- it knows everything about me, my work, every person I interact with, my goals, values and preferences
- I can interact via voice, text or phone
- output quality matches what I'd produce myself, or I can review and it's at least 10x faster

what can it do today? it reads and replies to any message (email, Telegram), does company research for investments, writes memos and reports, generates images in my style, manages my GTD system.

it has access to full transcripts of every call, every chat, even my philosophy and 5-10 year personal goals.

for people/market/company research, I built an orchestrator using MCP and tools (Perplexity, Parallel, Webfetch, Exa, Firecrawl) with depth levels. the deepest research runs for 60+ minutes and produces a small book.

let me show a concrete example I added this morning. I have a GTD system — basically just a file with a list of work tasks.

now I open Claude and type (or say) /gtd. then Claude:
- goes through the list and spawns sub-agents for each task
- prepares an execution plan per task (I adjust if needed)
- uses pre-built workflows for each task type (intros, podcast prep, trip planning, cold emails)
- has access to a global index of all calls, messages, companies and contacts for maximum context
- saves logs after execution
- self-improves once daily: reviews all corrections I made manually and adds them to the instructions

the screenshot shows two real tasks: one needed to google contacts and compose an email in my style, the other was deep research and preparing questions for a podcast. agents ran in parallel.

as you can see in the second screenshot, each task takes 15-30 tool calls and ~100k tokens on average, has an internal verification loop, but ultimately saves >24 hours per day. it's like hiring 5 clones of myself :)

soon it won't be 5. it'll be 5,000.

next step: roll this out to the whole company so we can all work on shared context and the clones aren't just mine, but belong to everyone on the team who's actually great at what they do.

[John Rush @johnrushx]

Sorry to be that guy,

but most vibe coders will soon realize that their business failures weren't due to a lack of coding skills.

[Gonka @gonka_ai]

No gatekeepers, no API keys that vanish overnight.

Run live decentralized inference.
Deploy through an OpenAI-compatible API.
Build intelligence that can’t be switched off.

Join the Gonka Discord to start:
➡️ https://t.co/0Q1izdpokj

[a16z crypto @a16zcrypto]

In this talk, @benleventhal shares his entrepreneurial journey of being the founder of Eater, Resy, and now @blackbird.

This conversation, moderated by a16z crypto general partner @AriannaSimpson, covers the evolution of going from a magazine to building a tech company, his advice for early stage founders, and how web3 enables Blackbird.

Check out the full talk here: https://t.co/OW0gosZxlK

You can also apply to our spring CSX cohort to hear these conversations in-person, plus off the record talks, workshops, office hours, & more: https://t.co/eUUUo6SovU

[Geoffrey Hinton @geoffreyhinton]

I was happy to add my name to this list of employees and alumni of AI companies. These signatories have better insight than almost anyone else into what is coming next with AI and we should heed their warnings.

[Elon Musk @elonmusk]

I do love this meme

[TON Community @toncommunity]

𝐓𝐨𝐝𝐚𝐲, 𝐓𝐎𝐍 𝐒𝐨𝐜𝐢𝐞𝐭𝐲 𝐥𝐚𝐮𝐧𝐜𝐡𝐞𝐬 𝐚 𝐟𝐢𝐫𝐬𝐭 𝐨𝐟 𝐢𝐭𝐬 𝐤𝐢𝐧𝐝 𝐢𝐧𝐢𝐭𝐢𝐚𝐭𝐢𝐯𝐞 𝐭𝐨 𝐮𝐫𝐠𝐞 𝐅𝐫𝐚𝐧𝐜𝐞 𝐭𝐨 #FREEDUROV.

An Open Letter Mini App in Telegram.

Anyone who values privacy and freedom of speech should sign this letter to French authorities. We hope to reach hundreds of millions of signatures.

We urge the French authorities to:
1️⃣ Free Mr. Pavel Durov from detention.
2️⃣ Allow Telegram to protect the freedom of expression and right to private life for its users.

The #DigitalResistance is now. #FREEDUROV.

🆓 𝐀𝐬 𝐚𝐥𝐰𝐚𝐲𝐬, 𝐬𝐡𝐚𝐫𝐞 𝐭𝐡𝐢𝐬 𝐩𝐨𝐬𝐭 𝐮𝐬𝐢𝐧𝐠 #FREEDUROV, 𝐚𝐧𝐝 𝐧𝐨𝐰 𝐬𝐢𝐠𝐧 𝐭𝐡𝐞 𝐎𝐩𝐞𝐧 𝐋𝐞𝐭𝐭𝐞𝐫 𝐢𝐧 𝐓𝐞𝐥𝐞𝐠𝐫𝐚𝐦: https://t.co/GRbGgGA8d8

[YAR @yar_chain]

Our world will never be the same. #Telegram gave us communication freedom. YAR supports a free world and totally supports #PavelDurov.
We ask all web3 world to support the #FREEDUROV movement and stop government censorship worldwide. Let's support projects that give us the freedom to be human.
#FreePavelDurov #freedurovnow

[Arianna Simpson @AriannaSimpson]

Two years ago, we co-led @morpholabs seed round; it's been incredible to see the progress that @PaulFrambot & the team have made over that time. We couldn't be more proud to be a part of today's Series A. Congrats to the team and onward!

[a16z crypto @a16zcrypto]

Here are some books we're reading and recommending this summer.

https://t.co/66DPPzjY7h

[Trader Tardigrade 🧬 @TATrader_Alan]

In this cycle, #Bitcoin is characterized with a down-sloping base in the Parabolic Curve Structure.🔥
We are at the BASE 3 and experiencing the same down-sloping moves like base 1 and 2.
Price action is consistent.
You can't chase it when it starts heading to base 4 🚀

[nick (Big Wick Nick) @ExitLiqCapital]

Linea ecosystem going BRRRRRRRRRR

$LINUS looking strong AF and still not even 1M

$NILE is my dex of choice on LINEA

$FOXY is the beta/proxy to metamask/consensys

next 6 months are gonna be lit

[Ferr @dontbuytops]

Metaverse & L1 runs were insane, right? But they won't be anything compared to AI & DePin runs.

There is are many catalysts coming that could form a bottom on every token and send them to new ATHs

Read this thread because i'll giving you:
- 4 catalysts that could boost every AI & DePin token
- 8 tokens that pull a 25-50x by EOY

Let's go 👇🏻

[CiPHERADX @cipheradx]

Since @cipheradx is an #adtech project with a #web3 heart, we should see #decentralized projects! Check "Friends With Benefits," a DAO like #tinder for #web3. 30K MAU and you never heard of it! Check it out at https://t.co/Wv3kTgtsih, and let me know what you think.

[vitalik.eth @VitalikButerin]

BInius: highly efficient proofs over binary fields

https://t.co/lHAGXbgI9r

[vitalik.eth @VitalikButerin]

And we finalized!

Happy merge all. This is a big moment for the Ethereum ecosystem. Everyone who helped make the merge happen should feel very proud today.

[Near Insider @near_insider]

ARAX LAND - A unique model that combines #Move2earn and #Play2Earn using #AR and #NFT technology. In the world of @araxland, you can find and collect mystical creatures called Araxes. 🎮

It is expected to be one of the hot games when released!

#NEAR $NEAR #Games