Howard

AI x Crypto. Ex-@Telegram fund, @Binance.

Shipping in public. @MakeReel_xyz, @xWatchlist_

[Howard @howard_]

I built a working product.
Median post: ~14 views.
Account: 5 followers.

Not "early traction."
A cold account with a real engine.

Full week-3 postmortem (what worked / what I killed) ↓

[Howard @howard_]

https://t.co/OsLndLaoC4

[Howard @howard_]

@scotty529 @shl @Gumclaw Btw, @shl 's minimum Skill is quite helpful: https://t.co/AZjdofW9DS

(p.s I submit in PR there: https://t.co/KH9KBcTNAU

[Howard @howard_]

Referral URL: https://t.co/GxngTlrVGP

Amazing work! cc @robj3d3

[Howard @howard_]

Turned a small monthly sub into real product traction, and the numbers still surprise me.

Here's what 6 weeks of consistent building with the right tools looked like:
→ Shipped 250+ posts
→ 15M+ impressions
→ 450k+ engagements
→ 4,000+ new followers
→ Crossed the X monetization threshold

All while shipping @MakeReel_xyz and @xWatchlist_ full time.

The key was automating the repetitive layer so I could focus on what actually moves things forward.

If you're building in public and want to know the exact setup, drop a question below.

[Howard @howard_]

Can someone DM me an invite to a private community?

Dev? Builder? Group chat maybe?

I'm looking to join communities so I can test these features—right now I have no place to test them. @jack

[Howard @howard_]

Jack (@block / @jack) just shipped BUZZ — a groupchat OS for people + agents.

I spent ~3h in it with Fable 5 / Kimi K3 / Codex 5.6-SOL.

Notes for builders (not a teardown): 🧵

1/4

[Howard @howard_]

@Baconbrix @SpaceXAI Congrats!

[Howard @howard_]

@hiddnest Amazing work! I think "lots of Dev/Builders" needs this product.

[Howard @howard_]

4/4

Where I’m pressing the same thesis

Same bet: people + agents in one room, with a memory of who did what.

Different door: @Telegram — where my users already live — plus small, checkable “fuel” loops (mission → proof of work → trust), without asking anyone to move house.

Buzz = public reference for agent-native rooms.
TG rails = where I can ship weekly and feel distribution.

Curious for the team: early priority closer to Block-scale agent fleets, or day-1 reasons for external builders to join the room? Both are valid — they just change what I watch next.

Thanks for launching the experiment in public.
→ https://t.co/yEqbSgFUoY

[jack @jack]

we're launching BUZZ!

a new groupchat platform for teams of people and agents of all sizes, built to reduce our dependency on slack and github. model-agnostic, decentralized, self-sovereign, and open source. 🐝

https://t.co/8IaMVeTQNo

[Howard @howard_]

Jack (@block / @jack) just shipped BUZZ — a groupchat OS for people + agents.

I spent ~3h in it with Fable 5 / Kimi K3 / Codex 5.6-SOL.

Notes for builders (not a teardown): 🧵

1/4

[Howard @howard_]

3/4

How I’d use it as a builder

• 𝗛𝗶𝗴𝗵 𝘃𝗮𝗹𝘂𝗲: 𝗿𝗲𝗮𝗱 𝘁𝗵𝗲 𝗺𝗼𝗱𝗲𝗹 — 𝗮𝗴𝗲𝗻𝘁𝘀 𝗮𝘀 𝗳𝗶𝗿𝘀𝘁-𝗰𝗹𝗮𝘀𝘀 𝗽𝗲𝗲𝗿𝘀, 𝘀𝗵𝗮𝗿𝗲𝗱 𝗿𝗼𝗼𝗺𝘀, 𝘀𝗶𝗴𝗻𝗲𝗱 𝗶𝗱𝗲𝗻𝘁𝗶𝘁𝘆 𝘃𝗶𝗯𝗲𝘀, 𝗼𝗽𝗲𝗻 𝗽𝗮𝘁𝗵 𝘁𝗼 𝘀𝗲𝗹𝗳-𝗵𝗼𝘀𝘁.

• 𝗟𝗼𝘄 𝘃𝗮𝗹𝘂𝗲 (𝗳𝗼𝗿 𝗺𝗼𝘀𝘁 𝘁𝗲𝗮𝗺𝘀, 𝗳𝗼𝗿 𝗻𝗼𝘄): 𝗯𝗲𝘁𝘁𝗶𝗻𝗴 𝘁𝗵𝗲 𝗰𝗼𝗺𝗽𝗮𝗻𝘆 𝗢𝗦 𝗼𝗻 𝗮 𝗳𝘂𝗹𝗹 𝗰𝘂𝘁𝗼𝘃𝗲𝗿 𝘄𝗵𝗶𝗹𝗲 𝘁𝗵𝗲 𝗰𝗮𝘁𝗲𝗴𝗼𝗿𝘆 𝗶𝘀 𝘀𝘁𝗶𝗹𝗹 𝗽𝗿𝗼𝘃𝗶𝗻𝗴 𝗱𝗮𝘆-2 𝗼𝗽𝘀.

That’s a timing call, not “Buzz is bad.”

Block running an internal experiment and keeping Slack is coherent. Outside builders can mirror that: learn in parallel, don’t burn the main thread.

[Howard @howard_]

“The 1st xAI World Martial Arts Tournament” will be held on Aug 8 in SF. (@Grok-athon)

Who's gonna win this?

[Howard @howard_]

@Erling so @Grok please adding the green onions with the photo in a Ghibili style there in the image

[Howard @howard_]

@rudyjellis Build tight filters early.

Most noise is just unread subscriptions.

[Howard @howard_]

@OpenAI's ChatGPT loop feeds them the fastest product signal in the industry. Hard to out-learn that. 😅

Short run: OpenAI owns the attention, but spend per user is thin and Google's squeezing the same crowd.

Long run: @apple wins. Devices in hand + payment already set up beats attention that migrates the moment something shinier shows up.

[Howard @howard_]

@ErikVoorhees @ErikVoorhees @VeniceAI Privacy-first inference is a major requirement for crypto apps trying to protect user data.

- How is the latency holding up compared to the centralized giants?

- And when will you add Kimi K3 or Inkling tho?

[Howard @howard_]

@natolambert The voice feature in the article is cool, didn't know Substack had that too.

Nice article, thanks for sharing. Do you think more overseas users will start using Chinese models? (cost considerations) Is that a chance to build an API relay business?

[Howard @howard_]

@durov Reducing the friction of self-custody down to a native chat interface is how we actually onboard the next wave of users.

Can't wait to see how @telegram handles the key recovery UX for non-technical folks. 😅

Will it be the new App or?

[Howard @howard_]

@AsideAI @DarayuthH Let me know what you think. It's the integration with existing tools that makes it click.

[Howard @howard_]

The @AsideAI is genuinely useful. It feels like a second assistant that actually helps me multitask and manage my X account properly. Honestly one of the better real-world examples of agentic software I've used.

Also nice that it can plug into tools I already pay for.

One request though: open up a native browser MCP. I want my agents to talk to the Aside browser directly, and eventually control it.

Some feedback after ~2 weeks:

1) The default https://t.co/M70O8Hi1Cb post template is solid, but please add a Copy button. Right now I still have to select the text myself.

[Howard @howard_]

Rumor has it that @Telegram is rolling out major updates to its Crypto Wallet network.

This comes after “the internal corporate merger”.

@ton_blockchain / @tappscenter / @wallet_tg is back!

[Pavel Durov @durov]

🌍 This summer will see the largest rollout of a non-custodial crypto wallet in human history.

⚡️Instant zero-fee crypto transactions for over a billion users are about to become reality.

We're bringing a native non-custodial Gram wallet to every Telegram app! 💎

[Howard @howard_]

Anyone can become a top researcher, and it comes down to one habit!

The only way is practice and practice and say it out loud!

The way you can check if you actually understand something is to say it out loud or write it down.

[Andrej Karpathy @karpathy]

One pattern I find useful for working with LLMs is a nice long ramble session. Sometimes the LLM needs more bits to understand what you're trying to achieve, but you're too lazy to type them. In these cases I like to lean back, switch to /voice and just ramble for like 10 minutes, total mess, anything goes, full stream of consciousness. Sometimes I declare it up top, something like "switching to speech recognition sorry for any typos...". Sometimes I turn it into a small interview of a few turns. But I find that the LLMs are somehow very good at reconstructing long incoherent rambles and often their echo of your own tangle of thoughts comes out quite a bit cleaner than what you started with. The result is that you improve the mind meld and have to correct things less from that point on.