Akshay Dodeja

"AI" in Logistics. It’s Already Here — Matt from @terminal49 and I discuss what the AI hype is all about it. We are at a technology inflection point in AI. We explore the significance of LLM's (large language models) and how the technology could accelerate innovation in logistics tech. Is AI just a bunch'a hype or is there potential?

Man this timeline is just getting weirder... Thats just based on stuff that is in the public domain.

Something is off about nvidia/jensen thing. You don't get fired for a top job for having an affair.

hey @vercel @rauchg $0.50/GB for log drains is goin to add up quickly. why so high?

This is interesting. For personal use, I use pond (pond.locker/reference/cli) I am building something like this for internal company use using pond + cloud storage + DIY MCP. I should probably jus open source it.

Fable 5.2: thick vibes so far Fast Responsive Efficient (at low) Easy going Eager Consuming

Happy Diwali 🪔

Happy Friday

Not many people talk about it but @claudeai Tag is a really good product. Slack is the control plane. ClaudeTag is realizing the true potential of @Slack.

Here is what I would do if I wanted to be the best CI provider: just for starters: Agent-first: CLI, API, MCP for everything JSON failures, not log dumps Logs agents can query Scoped, short-lived agent tokens Spend caps per agent Wait-until-done, no polling Reproduce any failure in a sandbox SSH into any job Drop-in for GitHub Actions One-line migration Sub-5s job starts No queues at peak No caps on job starts Every machine size: 2 to 64+ vCPU Arm, GPU, macOS, Windows Desktop-class CPUs Local NVMe No noisy neighbors Same speed every run Per-second billing No charge for queue time No charge for skipped jobs Best price per vCPU-minute Docker layer cache on the host Unlimited cache size Remote build cache Snapshots with services running Snapshots that rebuild on lockfile change Zero failures caused by the provider Network that behaves like GitHub's Fresh microVM per job Secrets safe from fork PRs Egress controls SOC 2, SSO, audit logs Test selection Shards split by test duration Flaky test quarantine Auto-bisect Fail fast Per-step CPU and memory graphs Cost per PR, team and agent Auto right-sizing Spot capacity with silent retry Built-in merge queue Preview environments per PR Run CI on uncommitted changes CI boxes as agent dev environments Agents that fix red builds Bring your own cloud Overflow from your own hardware Zero-config Rails, Node, Python Public benchmarks with variance Pricing without a sales call Support from engineers in minutes What did I miss?

We are spending ~$1k/m for CI on @useblacksmith. Tried out @boatdotdev to see how it compares. Much cheaper but not reliable.

Thinking of upgrading my son's Darth Vader costume for next year. Wouldn't it be cool if the chest control panel could control an actual "vocabulator" and change his voice? Here is a concept prototype. Try it with your voice. villain-voice-box.vercel.app…

Weren't there stories of Anthropic making $10B+ per month? $16B in H1 is much lower than that.

Been using @aquavoice for STT, and it's been working fine. I just tried Handy with parakeet model and it's so fast!

Token munching

abliterated models are causing a havoc and it seems likes going to get worse as these models improve weekly. 😰😰😰

Anyone concerned about the banking system? Most of it is not written on node or php? Right? RIGHT?

ouch internet

The AI labs moat is security vulnerabilities

What are the fastest github action runners? Currently using @useblacksmith but always looking for faster boxes