Ben Sparks

Smart Contract Lead @0xFastLane | Prev @Aave

[Ben Sparks @bensprks]

@keoneHD @category_xyz Does this mean staking epochs are shorter? ie still 50 000 blocks, but now takes ~4.17 hrs instead of 5.5 hrs?

[Ben Sparks @bensprks]

Levie’s takes on AI are worth paying attention to.

Slightly contrarian but well reasoned and directionally correct.

[Aaron Levie @levie]

We’re in a period where everything feels like it’s getting jumbled up across roles because AI lets you explore the adjacencies of other functions more easily.

We all collectively have to figure out the new form of definition of what these jobs look like in a world of agents, and certainly many will look different from what they did before. But there are some immutable laws that will eventually re-emerge over time and become clear again.

As an example, when you’re scaling, product managers should be spending an insane amount of time with customers and getting feedback on the product and thinking through what to do build next, how to design it so it’s usable, and so on. Engineers should be understanding the business objectives, and building systems that scale and are secure, even as feature velocity increases by 10X. Now both can do a bit more of the others role, and this can temporarily get conflated as doing the whole thing, but eventually the work adds up to be enough that it makes sense to specialize again.

Similarly, in GTM, the product marketer can certainly generate a working design and video for a launch, but the specialist is always going to (or should) have an eye for quality that delivers a better outcome.

My bet is that AI enhances specialization even further, even if a few roles collapse into each other, and the future toolchain and craft of the specialist will be much higher leverage and output far greater than anyone else as a hobbyist in that function.

[Ben Sparks @bensprks]

@mark_is_here @kartojal Mostly scheduled research reports via cron jobs, and like a more capable chatgpt with github access for light coding stuff. Still use codex for heavier stuff

[Ben Sparks @bensprks]

@kartojal @mark_is_here I use the same config, Hermes + Hetzner. Works great

[Ben Sparks @bensprks]

@mark_is_here +1 for Hermes on Hetzner. Never had problems with stability from it and @Teknium is shipping features at a ridiculous pace

[Ben Sparks @bensprks]

@hardtotelll Yeah, noticed my actual compute time of my Hermes agent on a VPS was minimal so thought a not-always-on agent could make sense

[Ben Sparks @bensprks]

Introducing Stratus☁️

An experimental, open source, Cloudflare-native personal AI agent. No always-on host required.

Inspired by @Teknium's excellent Hermes agent, and @Cloudflare's Agent Week announcements.

v0.1 includes basic memory, self improvement, tools, web browsing, 4 supported model providers, and more.

[Ben Sparks @bensprks]

Disclaimer: If you're looking for a reliable AI agent, just run Hermes in a VPS.

Stratus is still super early and experimental, an tbh mostly a way for me to play with Cloudflare's latest tech while learning about the internals of AI agents.

[Ben Sparks @bensprks]

Hermes' killer feature is self-improvement - your agent learns your preferences and gets better at the stuff you use it for over time.

Stratus includes an early attempt at a similar memory + self-improvement system:

[Ben Sparks @bensprks]

Hey @archethect the sc-auditor skill is awesome but can you rename the ATTACK phase - its scaring the models

[Ben Sparks @bensprks]

Hey @OfficialLoganK can we get a bigger Gemma 4? Would be cool to see what its capable of in Qwen3.5-397B-A17B size/config

[Ben Sparks @bensprks]

GPT 5.4 still mogs Opus 4.7 on code review

[Ben Sparks @bensprks]

Cloudflare is cool

[Ben Sparks @bensprks]

AI perf/cost barbell strategy:
- $100 Codex for most stuff
- GLM 5.1 via OpenCode for frontend/design stuff

Any reason to still have a Claude Code subscription I'm not thinking of?

[Design Arena @Designarena]

BREAKING: GLM 5.1 by @Zai_org overwhelmingly dominates design-centric coding tasks amongst open-weight models

In the categories featured below, it is most comparable to Opus 4.6 by @AnthropicAI at ~1/8th the cost

Huge congrats to the @Zai_org team for this achievement!

[Ben Sparks @bensprks]

Hermes + Qwen 3.6 Plus is a really effective agent combo

[Ben Sparks @bensprks]

@0xMimie gshmonad

[Ben Sparks @bensprks]

@blakeandersonw Took me 3 weeks just to get a personal developer account approved. They seem super overwhelmed with the influx of new AI-assisted devs.

[Ben Sparks @bensprks]

@devjoshstevens @Polymarket Congrats! Huge pick up for Polymarket