DigitalOcean

Most agents ship once and stay the same. The best ones learn from every run. Nov 12 in SF: happy hour with @NVIDIAAI, @sgl_project, and @MLHacks after SGLang Summit. Lightning talks and demos on building that loop. Request your spot: https://do.co/48pUak2

19 小时前 5 1

在 X 上查看 →

A looping script can spend a month’s AI budget in an afternoon. Billing Insights now breaks your spend down by model, token type and access key. Spend Alerts can pause a model access key at the budget you set, now in public preview. Billing Insights: https://do.co/3U4auUj

0:16

20 小时前 7 1

在 X 上查看 →

Thousands of first PRs, hundreds of events, one very good month. Proud to partner with @hacktoberfest . If you’ve been meaning to contribute, now's your chance. Find an event near you: https://hacktoberfest.com

1 天前 10 0

在 X 上查看 →

Models ship every ~10 days. GPUs turn over yearly. Fabs, every half-decade. @neimuc27’s advice: never hard-code the model name. Put it behind a router so a swap is a config change, not a deploy. Commit to compute on the slow clock, choose models on the fast one. ⏱️

1 天前 5 0

在 X 上查看 →

Join DigitalOcean in the Dev Zone at BoxWorks Dev Summit on Nov 5. On Nov 6, build with us in a hands-on workshop: an agent on Managed Agents reads a Box folder and saves its briefing back to Box. Free pass to both days: https://luma.com/box-dev-summit-2026?coupon=DIGITALOCEAN

1 天前 7 5

在 X 上查看 →

A new GPU can’t serve a token until the model loads. The article estimates a fully cold 70B start at 5–8 minutes. Warm nodes plus weight and engine caches cut it to 10–30 seconds, and engine startup becomes the bottleneck. https://do.co/3UT5X7p

0:09

2 天前 11 1

在 X 上查看 →

RT @dicabrejas: Coding assistants on your laptop are great, but they don't scale to a full backlog. If you want to move from single-prompt…

2 天前 0 3

在 X 上查看 →

DigitalOcean is on @Grok @bot. Sign in once, then ask in plain English to deploy to App Platform, launch Droplets, manage databases and DNS, or check what you’re spending. It shows a cost estimate before anything billable runs. https://x.ai/bot/plugin/79762613

0:15

2 天前 23 6

在 X 上查看 →

A system prompt is a request. A missing route is a guarantee. @neimuc27 ran an overnight support agent on 99 tickets: it could read, claim and draft, but had no way to send. All 35 sensitive tickets waited for a person. Full build: https://www.digitalocean.com/community/tutorials/ai-customer-support-agent-human-in-the-loop

2 天前 6 0

在 X 上查看 →

A restored agent microVM starts running before its memory is loaded. Pages come from storage only when the guest touches them. On the warm-template path, request to serving traffic takes under 200 ms at p50 in our internal testing. https://do.co/4soRAnA

0:17

3 天前 4 0

在 X 上查看 →

RT @paddix: You can now manage your @digitalocean infrastructure straight from @bot 11 skills for DigitalOcean, including product guidan…

3 天前 0 8

在 X 上查看 →

"Paste the API token into the prompt and hope” is not a security model. We let an agent run a real production migration across two clouds. It never held a credential, ran under an @NVIDIA OpenShell policy profile that locked writes to /workspace, and could only reach allowlisted

2:59

3 天前 8 0

在 X 上查看 →

A model’s suggestion to call a tool is a request, not authority. Action Gateway keeps credentials out of model context and checks every tool call against policy: allow, deny, or ask a human. It aligns with @nvidia’s Open Agent Safety Platform. https://do.co/4ierrTS

0:16

3 天前 9 1

在 X 上查看 →

Inference cost at scale is a routing problem before it’s a GPU problem. llm-d 1.0 went production-ready this month. It spreads token volume across the fleet to raise the cache-hit ratio. We run a forked version behind Serverless Inference and Managed Agents.

4 天前 10 1

在 X 上查看 →

.@MiniMax_AI reports 73.67% on DeepSWE and 72.5% on AutomationBench-AA, which it calls the best result among open models. M3.1 Flash is now on DigitalOcean Serverless Inference and Dedicated Inference. $0.15 input, $0.60 output per 1M tokens. https://do.co/4xsz0fI

0:06

7 天前 27 2

在 X 上查看 →

Over 90% of input tokens agents send us are tokens the model has already seen. Prompt Caching is now available on DigitalOcean Serverless Inference: cached input at up to 90% off, time to first token up to 80% faster, no code changes. Cache affinity keeps hit rates on flagship

0:16

7 天前 16 3

在 X 上查看 →

RT @paddix: This week’s PHP2Shell disclosures and the AI-assisted attacks involving ConnectWise show how quickly the security equation is c…

7 天前 0 2

在 X 上查看 →

RT @paddix: @JayaGup10 and @ashugarg make an important argument here: agents create an expanding execution workload around inference. My v…

7 天前 0 8

在 X 上查看 →

@PicapLatam @ProbablyDatabot @byteful hosts its website and dashboard on App Platform, with managed Postgres replicated near the edge and automatic failover behind it. https://www.digitalocean.com/customers/ping-proxies

7 天前 1 1

在 X 上查看 →

@PicapLatam @ProbablyDatabot ConfirmOk uses App Platform workers to place daily check-in calls to seniors and send alerts when someone doesn’t respond. Code deploys from GitHub, and Insights shows CPU, memory and request volume. https://www.digitalocean.com/customers/confirmok

7 天前 0 0

在 X 上查看 →

@PicapLatam @ProbablyDatabot, which builds AI agents, runs its agent orchestration, API and billing on App Platform. The first production API took a day and a half, and autoscaling absorbs bursty traffic from agents running on users’ machines. https://www.digitalocean.com/customers/probably

7 天前 1 0

在 X 上查看 →

.@PicapLatam runs ~70 apps on App Platform with one DevOps engineer. At 1.5M rides a month, with traffic peaks in the hundreds of servers, their CTO says scaling has been seamless. https://www.digitalocean.com/customers/picap

7 天前 2 0

在 X 上查看 →

App Platform runs your whole app: frontend, API, workers, scheduled jobs, and the database behind them, with AI a function call away. One bill, private networking between services and data. Successful teams building on it: https://do.co/4dQnhjN 🧵

7 天前 12 3

在 X 上查看 →

RT @ld_do_: Hey all, quick update: we tested a new UI with a small group of you, including a collapsible nav bar, and the feedback was loud…

8 天前 0 2

在 X 上查看 →