Jason Smale

i made this skill! one cool thing is that it matches frontier models to the type of task. in the skill rewrite, i opted to use Opus for prose and evals. for our backend cold start improvements, Opus served as subplanners, while GPT rapidly hillclimbed on perf

give an agent a fish and you feed him for a day teach an agent to fix ci and you feed him for a lifetime

@Itayd @Cursor we're constantly improving this but it still sucks when it happens ... got something for you to try. will dm.

Hey @Cursor is there a way for us to get BugBot to find all issues in one round? My PRs go through 5-10 rounds of back and forth with BugBot before it returns an all clear -- this is a real bottleneck. Love the product otherwise.

@__spongeboi @cursor_ai I haven't tried it, but a few lines in a bugbot.md might be fun to play around with.

. @cursor_ai i wish sometimes bugbot gave compliments , the constant critique is draining, but compliments all the time would also be bad. like once in a while, when a PR is absolute genius

We use a combination of offline benchmarks and online evals to measure model quality. This makes results more useful, especially as public benchmarks are increasingly saturated.

Automations with a full dev environment and MCPs configured make the dreamy future-of-work feel real. I’ve got ones for incident investigation, PRs for bugs in Linear, nightly code clean-up/incremental migrations. Built-in memory brings focus and continuity.

Story time! The very first version of Bugbot we launched was stateless. We kept track of runs in our analytics database and we only used that data offline for metrics. Whenever we needed to analyze run metadata, we had to do online lookups from the database. Relying on our analytics database during online operations was definitely tech debt. We were using a large RDS cluster that was already running hot from offline usage. We started noticing that some Bugbot runs were failing because of external analytics load; by that time, we had added a few Bugbot-specific tables to it, so we had to come up with a migration plan quickly before this became a widespread availability issue. We chose PlanetScale Postgres for our new Bugbot cluster. We already had many terabytes of data in PlanetScale MySQL/Vitess which was working incredibly well, so we felt comfortable with this choice. The migration went very smoothly, and now all of Bugbot's queries are instant and we stopped worrying about downtime!

I hadn't used Bugbot since launch. It's got a LOT better.

It’s official: McLaren F1 x Groq Bringing inference speed at a winning cost to the grid and beyond. See you in Singapore. 🧡🏁

Underdiscussed! As we push for more intelligence and lower latency in AI systems, serverside execution of tools becomes a more common pattern, and Responses API makes that possible in a clean way. Chat completions API, the legacy solution, has to be retrofit by each provider to make server side execution work and thus loses its main value proposition: standardization. This is just one of the benefits of responses over completions. Try the responses API on @GroqInc and let us know your thoughts!

It's over. OpenAI just crushed it. We have their o3-level open-source model running on @GroqInc at 500 tokens per second. Watch it build an entire SaaS app in just a few seconds. This is the new standard. Why the hell would you use anything else??

This is what 200 tokens/second looks like with Kimi K2 on @GroqInc For reference, Claude Sonnet-4 is usually delivered at ~60 TPS

*YOLO Launch* Kimi K2 is now in preview on GroqCloud at 185 tokens/sec. Build fast. Link in comments.

This solo founder built an open-source competitor to Perplexity with no team, no funding, and no permission. What started as a weekend project now powers over 1M searches, with 60K+ monthly users. This is the story of @sciraai and how Zaid Mukaddam built what others wouldn’t. 🧵👇

The @GroqInc team is rolling out some nice Llama 4 Maverick efficiency improvements to our global network. Check them out!

The Official Llama API ⚡️Accelerated by Groq In partnership with @AIatMeta. The fastest way to run Llama with no tradeoffs. Preview now live.

Llama 4 🔥

🔊🔛🔥 ... @GroqInc launched voice generation today. GroqCloud now has realtime transcription, LLMs, *and* text-to-speech. You can build super-responsive, ultra low-latency voice agents end-to-end entirely on Groq! At @trydaily, we're big fans of Groq's fast, low-latency inference. @pipecat_ai supports all the Groq models, including the new voice model. Lately, we've been obsessively playing a voice chat game that @mark_backman wrote. (My high score is 6, so far.) Here's Mark, with Groq's `Celeste-PlayAI` voice.

"We optimized our infrastructure to its limits – but the breakthrough came with GroqCloud™. Overnight, our chat speed surged 7.41x while costs fell by 89%. I was stunned. So, we tripled our token consumption. We simply can’t get enough." – Nicolas Bustamante, CEO, Fintool Read more in our latest customer use case with @fintoolx groq.com/groq-customer-…

The team is shipping at Groq speed⚡️ Tool use + JSON mode is up. We added a parameter to help parse the reasoning steps as well to make it easier to use