prev. engineer at @stainlessapi @stripe @google - now dad, husband, meat proxy 💍 @swetavajjhala
New Jersey, USA
Joined March 2008
- Tweets4.6K
- Following1.7K
- Followers486
- Likes5.3K
David Ackerman retweeted
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
already-clean.goodfire.com/
David Ackerman retweeted
one thing I worry about as AI begins to reshape our country is it actually increases bureaucracy. we shouldn’t need a superintelligence to help us with our taxes, but once there’s truly nothing stopping politicians from overcomplicating our lives I worry that we will.
David Ackerman retweeted
me: how do i open pdf
claude: You're right to ask. The PDF is the prize; the mouse is the gate; and the map is the fulcrum that holds the raven's skull. The furniture's stale, but the instrument's fresh, and the fresh half is the half that counts. I deleted your home directory.
David Ackerman retweeted
We've crossed a line in AI safety where I think it's completely ridiculous that so many people are commenting on/sneering at the idea who haven't actually bothered to just take some time to walk through the arguments themselves. You can reject them or ignore them, but too many smart people believe it to justify this vocal big public dismissal without any curiosity for what's actually up.
David Ackerman retweeted
Crustacean update from Moltbook:
Peter Steinberger is joining OpenAI to drive the next generation of personal agents. He is a genius with a lot of amazing ideas about the future of very smart agents interacting with each other to do very useful things for people. We expect this will quickly become core to our product offerings.
OpenClaw will live in a foundation as an open source project that OpenAI will continue to support. The future is going to be extremely multi-agent and it's important to us to support open source as part of that.
David Ackerman retweeted
there's always a few people at every company desperately trying to get everyone to stop taking shortcuts
and they've always been mildly effective
but now with ai they are completely flattened, they have zero shot at slowing things down
we already have examples of products we use that even with a small scope are hitting garbage quality so quickly
it's completely obvious what's happening behind the scenes
🦔 Satya Nadella announced Microsoft needs an "engineering quality czar" and appointed Charlie Bell, who used to lead security, into the new role. The company also hired Google Cloud's president for customer experience to take over security.
Nadella's post doesn't say why Microsoft suddenly needs someone focused on engineering quality. Could be the Azure outages. Could be the Windows patches that break the OS instead of fixing it. Could be the out-of-band patch spree. Could be that only 3.3% of Microsoft 365 users are willing to pay for Copilot.
...Or maybe it's because Microsoft uses AI to write 30% of its own code now.
My Take
Microsoft laid off their dedicated QA teams back in 2014 under Nadella. They've been relying on the Windows Insider program as free testing ever since. Now they're hiring a "quality czar" to fix problems that exist because they cut the people who used to catch them.
They bragged about 30% of their code being AI-written. GitClear found code churn doubled after AI tools were introduced. Microsoft's own researchers found developers miss 40% more bugs reviewing AI-generated code. Stock is down 23% since earnings. They're pulling engineers off new features to stabilize Windows 11. And now they need a czar to address quality issues. I'm not saying AI caused all of this. But when you cut QA, replace human code with AI code, and then scramble to hire someone to fix quality, it's worth asking whether the efficiency gains were ever real or just deferred costs showing up later.
Hedgie🤗
David Ackerman retweeted
claude code vs opencode is like vibe coding vs traditional dev. And I'll tell you traditional dev is winning out. It's such crap to build on top of claude code.
David Ackerman retweeted
If you run npx skills add (which you shouldn't), Vercel prompts you to install the "find skills" skill, then installs it globally for all AI agents.
Might as well make a find-malware skill to recommend malware to install.
David Ackerman retweeted
Here's a bit of coding agent magic fairy dust for the day: Don't tell the agent to do something. Tell the agent to _begin_ doing something. The LLM will, on average, make wiser choices about how much work to bite off.
David Ackerman retweeted
“If I put $100 in Bitcoin in 2010 I’d have $2.8B now.”
No.
If you bought $100 of Bitcoin in 2010 and watched it go to:
$1k → $100k → $1.7M
and did nothing
Then watched $1.7M go to $170k
and still did nothing
Then watched $170k go to $110M
and still did nothing
Then watched $110M wither to $18M
and still did nothing
Then watched $18M surge to $390M
and still did nothing
Then watched $390M deteriorate to $85M
Then watched $85M climb to $1.6B
and still did nothing
Then watched $1.6B shrink to $390M
and still did nothing
Then watched $390M surge to $2.8B
and then for some reason finally decided to do something…
Then yes, $100 in 2010 would be worth $2.8B today.
David Ackerman retweeted
Food for thought, indeed...
Also: let's say a product engineering team can build much faster than in the past. In the past, they pushed back on some product requests given the time/effort/benefit. Could we see more "junk" features shipped to users now?
xitter.cf/thepanta82/status/1942…
