I tested GPT-6.2 Astra on real work. 2 repos, 105 replanted bugs. Find and fix what you can.
It's the new #1. The results:
- GPT-6.2 Astra (max): 62.7 for $28.35
- GPT-6.2 Astra (xhigh): 60 for $23.53
- Fable 5.2 (max): 58 for $139.65
- Gemini 4 Carbon (high): 57.3 for $31.32
- Sonnet 5.5 (max): 51.3 for $154.03
- GPT-6 Astra (max): 45 for $33.03
From GPT-6 Astra to GPT-6.2 Astra: +17.7 points at max, +17 at xhigh.
Whatever people say about other use cases, it's a code review beast.
GPT-6.2 Sol and more effort levels drop soon in the thread 🧵
2/3
Buying a printer? Three rules:
* It has a LAN port (just in case)
* Wi-Fi MUST be configurable from the printer’s own screen
* It should list the available WiFi SSID's - if not leave it!
Ask the salesperson to connect it to Wi-Fi in front of you. No demo, no sale.
1/2
30+ years in IT, & I'm still humbled every time a customer asks me to get their printer “connected”, it should just work, it never does 🤦♂️
Easiest answer: a LAN cable from printer to modem/router.
Throw out the USB cable cause even the 90s don’t want that cable back.
Super excited about the Agent Board.
I've been struggling with multiple simultaneous sessions for months. The first idea was a kanban board, but that would create an even bigger mess. What seems to work: iitiatives with phases and open decisions on a single board.
Depending on your preference, different agents plan, build, and review (e.g. Claude, Codex, Grok). They wait for you or progress autonomously (next phase, release, etc.).
Soon in AFK Pilot (VS Code, Desktop App).
So I tested Fable 5.2 on real work. 2 repos, 105 bugs. Find and fix what you can.
The results:
- Fable 5.2 (max): 58 for $139.65
- Sonnet 5.5 (max): 51.3 for $154.03
- GPT-6 Astra (max): 45 for $33.03
- GPT-6.1 Sol (max): 44.3 for $5.88
- Fable 5.1 (max): 43 for $87.18 (n=1)
- Opus 5.5 (max): 41.7 for $58.53
All three Fable 5.2 runs beat Sonnet 5.5's average. And it got there with less than half of Sonnet's turns (707 per run vs 1,497), for less money.
More effort levels dropping in this thread soon 🧵
Pretty shit way to treat anyone - this whole “it wasn’t me it was the AI” is getting way out of hand - when are the laws/lawyers going to catch up and start getting punitive on this?
What actual f*ck @DarioAmodei / @boris / @TheAmolAvasare
Let me get this straight - I get this email telling me:
1. A cybersecurity "research" company gets into your internal systems while... *checks notes*
"conducting a security testing exercise on its own initiative and without our authorisation"
2. You "cannot confirm whether any particular person’s information was viewed," then in the same email say "we have no reason to believe you are at heightened risk."
3. You also try to reassure by saying "the cybersecurity firm has certified in writing that it has deleted the accessed data and will not use it."
Translation - they pinky promised they deleted everything they gained unauthorised access to.
I've been hacking companies for 15+ years, so let me spell it out.
Accessing systems without authorisation has a name -hacking. Calling it a "security testing exercise" doesn't change that.
Honestly, the access itself doesn't bother me much. The irony does.
I’ve spent over $50k on API credits in the last six months and contributing to CVP by accepting logging.
On top of that, I've reached out to your teams to collaborate on security research (actual research - not the kind that has you send out privacy notices to your customers) and heard nothing.
I'm getting deeply unserious security vibes and considering moving all my (and my clients) token spend to @OpenAI if I don't see material change here.
you'll likely have a real conversation with a whale in your lifetime
i mean literal back and forth, with AI turning your words into whale clicks
it's all thanks to Project CETI: the largest attempt in history to talk to another species
sperm whales have the biggest brain on earth
and they've been talking in the deep ocean for millions of years, but no human has ever understood a single word of it
so CETI trained AI on nearly 9,000 of their calls
and amazingly, it discovered a "phonetic alphabet" that's 7x more expressive than scientists thought
then they trained a language model on their actual whale conversations, which they call "ChatGPT 1 for sperm whale" lol
today, it can eavesdrop on them making plans and predict what they're about to do with 86% accuracy
for example:
before they go hunting, a family of sperm whales will float at the surface and click back and forth
then everyone joins in and they all head down together
it's basically a family deciding it's time to go get dinner
but we still don't know what most of those calls mean
and the way to find out is to start talking back
AI can already generate its own whale calls, so researchers can play them to real whales and see how they respond
every response teaches the AI a little more about what the calls mean
and the team behind it plans to make it conversational in the coming years
at that point, you say something and a whale answers
one researcher predicts the whale won't even be able to tell it's talking to a machine
that would be a world first: a real conversation with another species in its own language
and the one answering you has a brain 5x the size of yours
Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.
- Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all.
- Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract.
- Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code.
- And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.
The funnel is driven by a combination of factors:
1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel.
2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel.
3. Access. Free-tier, paid-tier, internal.
So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.
So, I finally tested GPT-6 Astra.
2 real repos, 105 bugs, find and fix what you can. Max effort:
GPT-6 Astra: 48/105
Fable 5.1: 43/105
GPT-5.6 Sol: 42/105
Gemini 3.8 Flash: 20/105
And it's crazy efficient. Look at the time and cost 🧵
A lesson from the Bug Hunt Bench: review your work with a different model. GTP-5.6 Sol (max) scored 1 point lower than Fable 5.1 (max), but found and fixed 8 bugs Fable missed.
this is the most Australian AI startup i've ever seen lol
a 19-year-old just raised $1.2M to build AI drone cowboys for cattle.
and the more i read about it the cooler it gets. here's how it works:
1. the rancher maps the whole property in the app, every fenced paddock, gate and water trough
2. they tap the paddock they want the cattle moved into, whichever one still has grass in it
3. the drone flies out autonomously, finds the herd, positions itself around the animals and starts guiding them toward the right gate
4. the AI watches their position, speed and posture, moving closer to apply pressure and backing off when the cattle show stress
cattle can hear the propellers from up to 100 meters away. when one needs an extra push the drone can direct the airflow onto its back
they even tested playing a recording of each farmer's own cattle call through the drone haha
and once they're through the gate the farmer gets a fresh report on the whole property, because the drone also:
> counted the herd on the way
> estimated what they weigh
> flagged any animal moving like it's sick
> measured how much grass is left in each paddock
> checked for open gates, dry troughs or busted fences
one drone can typically control and track herds of up to 2,000 animals. the company says it cuts mustering costs roughly in half
Brumby (formerly GrazeMate) is still rolling out pilots, but already had commitments covering 1.7 million acres across Queensland and New South Wales before expanding into California
so a job that used to require riders, motorbikes, horses or helicopters is now just 3 taps on a phone and an AI drone cowboy
i used to despise meetings
but i’m surprisingly starting to find them useful now
because agents are only as good as the context you give them, and a good 30-minute conversation can generate an *absurd* amount of it
> put a few smart people on a call
> work through a hard problem together
> uncover ideas none of you would’ve reached alone
> have wispr notetaker capture who said what, every decision, and every next step
> hand the full context to your agent
> start cooking the post-meeting work immediately
kinda counterintuitive but i honestly think AI might make the best companies have more meetings
because if execution is cheap and taste is the real bottleneck now...
then getting the right people together to decide what deserves to be built is one of the highest-leverage things a company can do.
Adobe has a pirate enemy.
He goes by comfyanonymous. Nobody knows his real name.
He built ComfyUI. The most powerful AI image tool on the planet. 123K stars. GPL-3.0. He ships a new release every single week. For free.
Then Adobe did the unthinkable.
March 13, 2026. The Department of Justice hit Adobe with a $150 million settlement for hiding cancellation fees and trapping millions of users in subscriptions they could not escape.
Then they raised prices again.
Photoshop is $22.99 a month. Illustrator is $22.99 a month. Premiere is $22.99 a month. Photography Plan is $19.99 a month. Midjourney Standard is $30 a month. Firefly locks every real generation behind a credit meter that resets every month and never rolls over.
The trillion-dollar creative empire got caught red-handed by the DOJ, then squeezed harder. They called it a "value adjustment."
Coincidence.
Here's the wildest part:
comfyanonymous did not slow down. Latest release: August 3, 2026.
Still free. Still GPL-3.0. Still runs on your own laptop. No credits. No cancellation trap. No monthly rent. No cloud upload of your work.
Same models. Same power. Zero dollars. Forever.
One anonymous developer vs. the trillion-dollar creative empire.
But DO NOT install it. We should all keep paying Adobe.
100% Open Source.
(Link in the comments)