Tag
#agentic
9 entries tagged "agentic".
-
Grok Bot feels like Claude Code did last year
Claude Code was the first harness that made agentic work feel shippable. Grok Bot is the first cloud-native personal-assistant stack that gave me that same jolt: named agents with real computers, and roughly zero ops.
-
Same harness, more engines: when one model stops being the default brain
I was an Opus fan. Then Opus 5 spent a session proving a solvable problem impossible, and Fable answered every question with a wall of tokens. So I stopped routing every hard job to Anthropic and started matching the model to the task. Grok shipped what Opus explained away; GLM does the cheap bulk. The skill I actually needed was routing.
-
1 token on your wrist: a $40 Claude meter
A $40 ESP32 smartwatch reads my Claude quota in real time. The trick is that the watch never talks to Claude. An OCI VM pings the API every five minutes with max_tokens 1, reads the quota out of the response headers, and pushes the result through Tailscale, an Android foreground service, and BLE to a wrist.
-
June: Scaling per-team environments, the DevSecOps loop, and the "living knowledge" decay
Three threads from the month. Migrating AWS dev environments to a DRY Terragrunt root, building a shift-left DevSecOps loop to triage GHAS and Black Duck noise, and the existential risk of abstracting the terminal away from engineers.
-
A town of $0 AI characters, built while I slept
DevOps in 2026 is also FinOps for AI. Following OpenClaw, my $0/month personal assistant, Holtwick is a town of NPCs at $0/month: twelve persona-driven AI characters whose voices stream in under a second from Cloudflare and Groq's free tiers, in a 3D world that overnight Ralph loops built while I slept.
-
OpenClaw on Oracle Cloud Free Tier: Lobster in the shell
An AI assistant on $0/month infrastructure: open-source middleware on a free cloud VM that, once you live with it, looks suspiciously like the core of the AI hype itself.
-
May: AI on both sides of the diff, browser tests as state guardrails, MCP as production integrations
Three threads from the month. The honest tension of using one AI vendor for both codegen and review, browser-test-as-code (Playwright at PR time plus Checkly and Datadog as synthetic monitoring) becoming the practical guardrail against AI-introduced regressions in stateful systems, and MCP servers maturing into the agent runtime's SDK layer without the operational discipline that usually carries.
-
April: Agentic engineering in production, AWS wired in Terraform, benchmarking SCA tools
Three threads from the month. The agentic workflow stack moving from sketch to production (gh-aw source/lock split, MCP versioning, network firewall), AWS dev environments wired end-to-end in Terraform with Fargate Spot pulling most of the cost out, and a tour through the SCA tool market that turned into a complaint about how vendors run technical evaluations.
-
March: PKI before the mid-2026 deadline, an agentic engineering setup, and infra hygiene at scale
Three threads from the month. The Chrome root program splitting serverAuth from clientAuth and what that means for internal authentication, formalizing how I run multiple coding agents in parallel without them stepping on each other, and a sweep through the boring infra hygiene that quietly stops working at scale.