Applied AI Engineer | @cognition and @n8n_io Ambassador | Growth & Product | Data-Driven CX
Mars, Planet
Joined April 2026
- Tweets2.2K
- Following737
- Followers251
- Likes3.6K
this is the part of AI i actually care about
coding agents are fun, but a model helping find new medicines is the real benchmark
Over the past few months, Anthropic scientists have engaged in deep research partnerships with six organizations developing new medicines. We've also begun drug discovery efforts of our own.
In doing so, we’ve learned much more about how Claude can support the many stages of drug development.
In this blog post, we share some early results from these efforts, and describe how this work fits in with our wider ambitions in the life sciences.
anthropic.com/research/early…
Fanfulla retweeted
The bar for quality in FC 2.0 is very high
We rejected 80% of the PRs submitted and sent them back for OSS maintainers to improve
This is now the eval that best matches my vibe check of models
Introducing FrontierCode 2.0: our new coding benchmark that measures a model’s ability to go beyond the spec.
Most coding evals measure a model’s ability to write code that meets a spec. With FrontierCode 1.0, we asked a new question: would you actually merge the PR?
FrontierCode 2.0 is comprised of more diverse tasks, featuring larger scopes and higher mergeability standards. We are now asking: not only would you merge this PR, but should you?
One key feature of FrontierCode 2.0 is underspecified prompts. Rather than including every engineering decision, prompts are deliberately ambiguous, giving the models room to demonstrate their judgment.
The result is a benchmark that is far from saturated, with the top model scoring just 45%.
well it's honestly true even at the scale of my own agent setup
it got better when i stopped swapping models and fixed who checks whose work
I think it's vastly underrated how important org chart and team functionality is to building a good LLM. Over 18 months old, and still one of my favorite all time Interconnects posts. It's why we still often see orgs crumble and need to get rebuilt (e.g. Llama/xAI) and why I believe underdogs can still catch up. Having an effective org and team is worth multiple multiples in compute.
I also think there isn't that much secret sauce to building models. Yes, sure there are some tricks, but most of the gains come from consistent execution and using your compute wisely.
Fanfulla retweeted
12 months ago 100% of my coding was done locally against my computer's filesystem
now ~5% is done locally, my computer is a conduit to download and upload files to cloud agents
most days my iPad and iPhone are all I need and I'm getting 50x more work done than I was a year ago
Real NPCs that talk and react in real time
we're one update away from NPCs with better small talk than us lol
The reason i'm leaning toward the Open Weights model isn't only the price (it has a big part tho)
it's that the copy on my disk can't get quietly worse next week
I'll fold most subagents but i'll keep the reviewer separate just in case
Otherwise the same model is grading its own diff
Before adding another subagent ask what problem it's solving. A bloated system prompt is often the issue.
Skills and tool search can fix that without splitting context across agents. My rule of thumb is to keep a subagent when the work is parallel, needs fresh context or one agent really can't pick the right tools.
Otherwise, fold it into the main loop, turn its prompt into a skill, run your evals and keep whatever wins.
Good write-up:
claude.com/resources/article…
Apparently now even "--dangerously-skip-permissions" doesn't allow "rm -rf ~"
Now there's an .env var to fire the last adult in the room.
Fanfulla retweeted
big Grok Bot Shelf day 📚
~70 new bots live on grokbotshelf.com — builders shipping real tools all morning
garage · fed decoder · heimdall · groktext · AI usage hub · box solution builder · shelfer · and a ton more
huge thanks to everyone who got shelved today:
@OutworkVentures @weaverwork @howdyollie @nahjuin @thepetermick @phillipggreen @EmilioSchwaiger @josephstayhappy @_h4xin @aceop_xyz @ALL_Jonah @CaliLumberJack @chaddsmith @cillfore @DarthJML @dbarrett24 @Fanfulladev @HudsonSutherla2 @HungOnCode @iLytzki @itsryanlenk @JayinEly @JeebsTX @JennaLiftsLife @JimothyRac @Jorhan @magnus72 @moondough @prince_twets @richphotoio @ryangrowthh @sommelierbot @sproutmed @StealthTSLA
want on the shelf? reply with your x.ai/bot/ link or tag me or @thecsguy
I'm really not understanding OpenAI.
They just released GPT-6.1 Terra and it lands well behind Opus 5.5 in Claude Code on the Coding Agent Index.
What's the point?
Another big drop from Cognition!
Devin went from writing secure code to securing the code you already shipped, with Mythos doing the checking.
No other ADE is shipping at this pace right now
The world's most powerful cyber models, including Claude Mythos, are available in Devin Security Platform. We're also offering hundreds of millions of dollars in credits to use it.
If you want to defend faster than attackers, you have to find and fix flaws at machine speed. That’s why we’re building the best AI-powered security platform in the world.
Devin already helps engineering teams write more secure code, now it can secure the code they already have.
Fanfulla retweeted
The world's most powerful cyber models, including Claude Mythos, are available in Devin Security Platform. We're also offering hundreds of millions of dollars in credits to use it.
If you want to defend faster than attackers, you have to find and fix flaws at machine speed. That’s why we’re building the best AI-powered security platform in the world.
Devin already helps engineering teams write more secure code, now it can secure the code they already have.
It's time to touch some grass (or iron 🏋️)
Summer bodies are built in winter, so I built a personal trainer Grok Bot:
- weekly plan from the equipment you actually have
- live coaching at the gym with exercise videos
- every session logged + progress dashboard
- a message 30 min before each workout, from your calendar
Skipping leg day now means ghosting my own bot
Add it to your Grok Bot: x.ai/bot/yLMfIOV5mVCoPL_4UPF…
Exercise clips: FitnessScape & Andrew Kwong (CC BY 3.0), Taco Fleur (CC BY-SA 4.0) via Wikimedia Commons, modified. Music: "Melodic EDM Loops" by Fupi (CC0). This video: CC BY-SA 4.0.
Anthropic is filing for its IPO today
The first detail i want to check in the S-1 registration statement is the actual number of tokens allowed under the weekly limit lol
Benchmark news
Gemini 4 Argon jumped on FrontierCS 2 just by switching harness and a chunk of its runs ended on provider refusals (classic)
Before I blame my harness again, i'm counting how many of my failed runs were just refusals
gstack, pstack, and now tstack.
If this keeps going I'll have to publish fstack, which is two skills and a note that says run the tests first
Rumors about Fable 5.5 that is already running at JP Morgan and Meta (i wonder why they don't use Muse 🥁)
So next week's "launch" is really just us getting Fable 5.5 after the banks already had it
Finally Kimi apologized for its rate limits and gave everyone part of the month back ( i wish they did that too when Kimi K3 came out too)
Just forwarding this to Anthropic for when my weekly limit dies on Thursday
We're sorry for the rate limits and unnecessary quota consumption on the K3.1 Pro subscription plan. These were our issues, and we should have fixed them sooner.
Today, every subscriber on every plan gets a one-time usage reset, restoring 25% of their monthly quota (usage below 25% resets to zero), equal to one week of usage. Your usage bar will reflect this adjustment by end of day PST.
We've also adjusted the plan's parameters and are reducing unnecessary subagent calls and output length, so your quota goes further.
Thank you for your patience while we put this right.