Principal Infrastructure Engineer at Splunk; ex AWS, Cisco; speaker; content creator
Krakow, Poland
Joined October 2011
- Tweets7.8K
- Following904
- Followers9.2K
- Likes11.7K
Dmitry Figol retweeted
extremely unprofessional. if kimi wants to make it as a frontier lab, they need to act like one: perhaps silently route people to worse models, and maybe write a blog post about the collapse of humanity
Kimi K3 has received far more love than we expected, and our GPUs are feeling it.
Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritizing compute for current members. Existing subscribed users are not affected.
We're adding capacity as fast as we can and will reopen new subscription spots in batches.
Going forward, we'll also split membership into two more focused plans: Kimi Membership for Kimi Web, App, and Work; and Kimi Code Membership for coding workflows. This will help us match compute more precisely and keep the experience stable.
Thank you for your patience and understanding!
Dmitry Figol retweeted
Trading can be incredibly challenging. People rarely discuss the tough moments or the sleepless nights filled with doubt.
There are times when it feels like all your efforts, time, and resources have been squandered. Moments when you’re alone, grappling with the heavy question: “Is trading really going to pan out for me?” Days when you put in the work but see no rewards. Periods when you reflect on every sacrifice made to pursue this journey. Instances where you second-guess your approach, your methods, and even your own abilities.
These struggles are genuine and unavoidable—you’ll face them. If you’re in the midst of them now, here’s my simple advice: Keep going. Persevere. Never quit. You’ve invested too much to throw in the towel. You’re nearer to unlocking your full capabilities than you realize. Hang in there—it’ll all be worth it in the end.
$SPY $QQQ $CRWV
Dmitry Figol retweeted
If you don't already know, it's clear how this will all end.
Anthropic will make legal moves against incorporated entities (like open code) and send C&Ds to OSS projects like clawdbot or charm
But it won't matter, the auth is still client-side!
Dmitry Figol retweeted
I'm Boris and I created Claude Code. Lots of people have asked how I use Claude Code, so I wanted to show off my setup a bit.
My setup might be surprisingly vanilla! Claude Code works great out of the box, so I personally don't customize it much. There is no one correct way to use Claude Code: we intentionally build it in a way that you can use it, customize it, and hack it however you like. Each person on the Claude Code team uses it very differently.
So, here goes.
Dmitry Figol retweeted
It's been years - back to Twitch to repeat a session that many attendees said was their favorite of re:Invent. We'll do an interactive deep dive into emergent behavior, metastable failures, and how to prevent long outages. This Friday, 10AM Pacific time. twitch.tv/colmmacc
Dmitry Figol retweeted
LLMs are already weird black boxes with unpredictable behavior
almost every coding agent going around them is closed source with the approach of "trust us we know better than you"
except they don't, can't beat a million devs who can dig into how the tool works and collaborate
wow ok i think i just hit the point that will make me stick to @opencode:
Open source means the model processing is not a black box anymore! You can check how many lines the agent reads when it looks at a file, why it stops the shell command, etc. You can know what's going on!
Dmitry Figol retweeted
Our second Online keynote is live from @thdxr of @OpenCode, arguing that the real hard problems of building AI products have not changed:
youtu.be/o3gmwzo-Mik
While AI can generate endless slop, it cannot generate cool and "aha" moments. This requires deep taste, not mindless optimization. Here is how he has applied it for SST/OpenCode!
Dmitry Figol retweeted
>see new codegen model
>ask company if model is fully custom or just post-trained on chinese oss
>they don't answer
>explain the differences between fully custom and just post-training
>they laugh and say "it's a good codegen model sir"
>try the model
>it's chinese
Dmitry Figol retweeted
today we're launching opencode zen
zen provides the best LLMs for coding ALWAYS in the best possible configuration
no more "did claude get dumber?"
no more being routed to a crappy provider
and we're giving it to you at cost
and you can use it anywhere
that enough for you?
Dmitry Figol retweeted
Almost 4 months later, here it finally is:
uv: Making Python Local Workflows Fast and Boring in 2025
youtube.com/watch?v=TiBIjouD… 1/3
Dmitry Figol retweeted
Seeing the cursor thing go down, it sort of validates the need in the market for tools that make more efficient use of tokens.
Claude can burn tokens for now because of max pricing, but everything else is fighting gravity - being efficient matters a lot!
Dmitry Figol retweeted
Something I’m incredibly proud of is finally live.
We've launched the Cloud Security Championship: a 12-month series of deep-dive challenges, each crafted by a different top Wiz researcher.
The first challenge is up. Go!
cloudsecuritychampionship.co…
Dmitry Figol retweeted
Cloudflare Containers are now available in public beta. Deploy simple, global, and programmable containers alongside your Workers: cfl.re/4kNjs06
Dmitry Figol retweeted
Amusing outcome of "vibe coding" wave with AI (by people who have not written software professionally before)
Seeing the speedrun of why software development REALLY is hard
(it was never about writing code - even though most non-dev folks assumed it must have been!)
Dmitry Figol retweeted
years ago i had a mindset that most engineers have
i believed if i produced a good product then marketing was not necessary
my bar was very high - good meant people loved it and were telling their friends about it
this isn't completely wrong but if you visualize a funnel this does not consider the absolute top
why would someone going about their day drop everything and try my product long enough that they hit the aha moment
what made it click for me is this video by veritasium (link in reply)
he's a youtuber that makes genuinely great content. it's not debatable and very few people produce anything at that level
he talks about how much effort he puts into packaging his work and thinking about how to get people excited to watch it
seeing someone that good still have to consider these things made it really clear what was going on
i was afraid of marketing, an area that was unfamiliar and unnatural, so i adopted a pretentious view of things to avoid confronting that fact
i'm on the other side of this journey now and it's painfully obvious how the majority of people are stuck in the exact same place i was
Dmitry Figol retweeted
ty, our upcoming static type checker and language server for Python, is accidentally on the front page of HN.
We're rapidly closing in on an initial "experimental preview release"...
Dmitry Figol retweeted
Yup. Assignments should be like “use AI to do ___ complex project” and then have an oral component where the student explains each step of the process they used and their thought process behind it. Schools should prepare kids to accomplish things with the tools of their world.
The solution is to embrace LLMs, encourage students to use them as "thinking buddies" or private tutors at home––which is how I use it, and how they'll be using it for the rest of their lives. For exams, pivot to ~100% in-class essays and oral exams. Worked like a charm in my University of Austin seminar.
I'm not sure this is a net-bad change over the status quo. And even if it is, it's inevitable anyway. Resisting it is like resisting the advent of calculators.
The alternative route is a stupid & unwinable arms race between cheaters and cheat-detection. Professors are destined to lose this arms race, b/c the incentives to successfully cheat will always be stronger than the incentive to catch cheaters.