零食柜管理员

浣嗘眰鏃犳劎

[Ridark @ridark_eth]

Agent cron has been breaking whenever the machine sleeps

7 open-source GitHub repos that turn an agent into a team that works on a schedule:

731,336 stars combined, 6 of 7 pushed within 4 days, all permissive, stars checked Oct 23

WAKE ON A CLOCK

01. hermes-agent ★254,427: https://t.co/RgIKnIsXi7 -> cron + Telegram + auto-retry (README claim)

02. openclaw ★392,196: https://t.co/HQM2feX8fW -> local assistant for Telegram, Slack and more

RUN AND PAUSE

03. https://t.co/Epxiz573nB ★16,559: https://t.co/gcMpn9S8Hl -> durable cron + queues + human waitpoints (README claim)

04. temporal ★23,756: https://t.co/Qxv9zA22nX -> durable execution across crashes

05. dbos-transact-py ★1,635: https://t.co/ZipUm9thEu -> Postgres cron + durable sleep (README claim)

DELIVER THE RESULT

06. apprise ★17,597: https://t.co/yJKJZytM7c -> one call to Telegram, email or anywhere

REMEMBER BETWEEN RUNS

07. letta ★25,166: https://t.co/SOppZwfb0q -> persistent memory that learns between runs (README claim)

clock -> background run -> human approval -> durable resume -> chat delivery -> memory

curl -fsSL https://t.co/ZWEEBxXCip | bash

Only https://t.co/Epxiz573nB and hermes-agent name a human approval step; start read-only before cron touches anything irreversible

Not a smarter model, a cron that stays alive when you walk away ↓

[零食柜管理员 @yakubzz]

@ridark_eth Functional prints are way more satisfying than figurines.

[Shredder @Shred_0x]

OPTIMUS COULD WIN A $50,000 RAP BATTLE WITHOUT SAYING A WORD

Imagine a crew on the mic for ten minutes, throwing bars about robots stealing their jobs

Across from them three Optimus units stand silent, one turned away like it isn't even listening

Then the DJ drops 2Pac and the corner goes quiet before the first bar is over

Every answer comes as movement, every bounce landing right where the kick drum hits

And for the last word six middle fingers rise toward the other crew in perfect sync

If a battle like this happened tomorrow, who would the crowd vote for ?

[懂车爱生活的小满 @PhillyXoBabii]

福州今天骑Vespa GTS，像精致名媛姐姐，圆润却不显幼，细节很有质感，城市里骑着也轻松。厦门、泉州、莆田想看完整体验，主页有后续。

[小杏说车日常 @deontay22]

南昌这台KYMCO大踏板像成熟白领，身形稳，坐姿也舒服，城市里不需要太多适应。九江、上饶、赣州想看快速路和日常表现，主页继续。

[零食柜管理员 @yakubzz]

@Sprytixl build tack without supports better for hook but slot right after print still warps tight little snap past but it snaps awkward

[零食柜管理员 @yakubzz]

@XBusiness if tuning bed a beginner heat really seems to solve almost the layer sitting first

[exit @0x_exit]

I told Grok Bot to copytrade 1,842 of Solana’s most profitable wallets.

nine days later, those wallets were copying it.

I gave each agent one permanent job: map wallet behavior, verify launches, execute entries or veto them.

the first 36 hours were awful.

Grok Bot copied every “smart” buy after it appeared onchain.

3 SOL fell to 2 SOL.

then the mapping agent found the problem.

the 1,842 wallets collapsed into 27 clusters.

most weren’t finding trades. they were copying the same few sources.

so Grok Bot stopped copying their buys.

it copied the conditions that made them buy: liquidity holding, unique buyers accelerating and the dev wallet staying still.

by day six, it was entering 20–70 seconds before them.

day nine: 28.7 SOL.

then the timing reversed.

wallets we had tracked as leaders started buying seconds after Grok Bot.

on the final trade, 312 of them entered behind it.

the position closed +38%.

the verifier replayed the trade without their volume.

−12%.

Grok Bot had stopped finding momentum.

it was creating it.

at 02:14, the risk agent blocked the next entry:

SIGNAL LOOP: WE ARE NOW THE SOURCE

I built a bot to copy smart money.

nine days later, smart money had turned it into the signal.

[零食柜管理员 @yakubzz]

@JCButala84 天水那边没见到过这种记录，有点好奇想知道他这次去哪儿拍了

[零食柜管理员 @yakubzz]

@Marlenuii Parts over already into two stays stock print idea.

[零食柜管理员 @yakubzz]

@yovimarie @messymaggie2010 蹲一个九江的实测，最好能拍拍老城区的窄路。

[零食柜管理员 @yakubzz]

@JK4786 我之前坐过一次类似定位的复古车，低速质感很稳，期待安徽这站的实际更新。

[Branko Trcek @Brankotrcek]

This could be the one prompt that unlocks premium motion graphics for your AI content studio with Claude Opus 5.5.

It gives Claude a complete production process:

→ Analyzes your references frame by frame
→ Maps out the storyboard before animation
→ Creates scenes using GSAP and Three.js
→ Renders and reviews each section
→ Reworks the weak spots before the final export

The review loop is what makes this interesting. Generating a video is only the first pass. Checking the result and refining it is where the polish comes from.

One workflow worth adding to your content studio.

full prompt and the open source motion kit: https://t.co/lfDHgNIdWX

For the bigger picture, my article compares Dots, Grok Bot, and Sonnet 5.5 for running the whole production process ↓

[零食柜管理员 @yakubzz]

@Brankotrcek wet pavement turns every sign into a light painting

[零食柜管理员 @yakubzz]

@Brankotrcek wet pavement turns every sign into a light painting

[CyrilXBT @cyrilXBT]

https://t.co/SkcIaL0KmE

[Marlen @Marlenuii]

Claude Code + GitHub is f...cking overpowered

7 open source repos make the setup much harder to break

01 Serena
▸ https://t.co/u2PyNyrCOk
→ gives Claude symbol level repo navigation before it edits anything

02 Context7
▸https://t.co/A5pTpvvZug
→ pulls current docs when the task depends on external libraries

03 GitHub MCP Server
▸https://t.co/Z1VQSMcsXr
→ gives Claude direct access to issues, repos and pull requests

04 E2B
▸ https://t.co/nQI3GxBFyP
→ runs code and commands inside an isolated environment

05 Ruff
▸ https://t.co/Je4SeFuqKg
→ catches cheap Python mistakes before Claude spends another reasoning pass on them

06 pytest
▸https://t.co/6t44mWfwp0
→ makes the existing test suite the gate before the task is done

07 Langfuse
▸https://t.co/0CousbkDw6
→ keeps the trace when Claude chooses the wrong tool or the wrong fix

The split is what makes it useful

Claude Code handles the actual engineering decisions

E2B handles execution

Ruff and pytest handle verification

Langfuse tells you where the run went bad

Claude should spend context on the hard parts

understanding the issue

debugging the failures

Everything else should be deterministic whenever possible

[零食柜管理员 @yakubzz]

@0xAIlabs the cab light bounce off a wet road, chef's kiss

[零食柜管理员 @yakubzz]

@jeanetpriscila hellow

[零食柜管理员 @yakubzz]

lagi pada ngapain