Zack Swafford (@zswaff) header
Zack Swafford (@zswaff) profile picture

building https://nitter.gq/t.co/qa2IRUHyVW (YC W22), where humans and AI agents actually get work done. wiring agents into real teams and posting what breaks.

San Francisco, CA
Joined March 2012
Photo by Zack Swafford (@zswaff)Photo by Zack Swafford (@zswaff)Photo by Zack Swafford (@zswaff)Photo by Zack Swafford (@zswaff)Photo by Zack Swafford (@zswaff)Photo by Zack Swafford (@zswaff)Photo by Zack Swafford (@zswaff)Photo by Zack Swafford (@zswaff)Photo by Zack Swafford (@zswaff)Photo by Zack Swafford (@zswaff)
My cofounder Milad and I started Dart because we both saw the same thing coming, independently, at different jobs: AI was about to transform every vertical SaaS industry. And project management was going to be one of the first to change. Here's the story:
3
1
19
2,163
Zack Swafford retweeted
Dart docs can be public now. Click the share icon at the top of a doc and pick Public. You get a link anyone can open without a Dart account, and they see the doc read-only. Tasks and other docs linked inside it still only show up for your teammates.
1
29
people really forgot google basically invented half this stuff and has unlimited compute sitting right there. feels like we're finally seeing what happens when deepmind puts it all together
1
1
23
tied on score, neon wins on speed and astra wins on api cost. tough chart for fable at $139 though
Google cooked. I tested Gemini 4 Neon on real work. 2 repos, 105 replanted bugs. Find and fix what you can. It's the new #1. The results: - Gemini 4 Neon (xhigh): 63 - GPT-6.2 Astra (max): 62.7 - GPT-6.2 Astra (xhigh): 60.2 - Fable 5.2 (max): 58 - Gemini 4 Carbon (high): 57.3 - Gemini 4 Neon (high): 57 - Gemini 4 Carbon (medium): 56 More effort levels for Neon (medium, low) are dropping in the thread soon 🧵
1
36
love that the harness efficiency dip barely lasted a month before spend doubled again
The 90th percentile OpenAI researcher now uses around $23,000 worth of tokens per day, up 105% from September.
2
29
love that ai vagueposting has officially made it all the way up to the c-suite
1
5
240
at least now when you hit the usage cap you can tell yourself you're a shareholder
BREAKING: Anthropic's IPO has now attracted nearly $500 billion in investor orders, per CNBC. Institutional demand has reached ~$400 billion, while retail orders are approaching $100 billion. The company is now considering expanding its IPO and pricing shares above $80, above the initial range of $70 to $78 per share. Anthropic is also considering raising more than the initially planned $64.4 billion.
1
3
84
48 hour ping time
IT’S OFFICIAL!! VOYAGER 1 HAS BECOME THE FIRST HUMAN-MADE OBJECT TO REACH A DISTANCE OF ONE FULL LIGHT-DAY FROM EARTH!!! NEARLY HALF A CENTURY IN THE MAKING.
1
88
nobody is working tomorrow. 99 for gta 6.
1
1
162
$200 for a java zero day
BREAKING: A cybersecurity firm used AI to uncover a critical security flaw in Java, one of the world’s most widely used programming languages. The bug, dubbed “Java4Shell,” was found by start-up Calif in OpenJDK, the core software that runs Java programs. Researcher Quang Luong used OpenAI’s Codex coding agent to find the flaw in about an hour, then build a working attack in about two more, for less than $200 in computing costs. Calif rated it 9.2 out of 10 for severity. The flaw sits in Java’s Base64 decoder and can be triggered by sending just 64 “A” characters followed by a single “=.” That can corrupt the computer’s memory, potentially letting an attacker take over a server or expose sensitive data, similar to the Heartbleed bug. Calif demonstrated a working break-in on Keycloak, a widely used login system, and says tools like Jira, Confluence and Jenkins are also at risk. Java runs at 90% of Fortune 500 companies, Calif estimates, and is used by banks, hospitals and governments worldwide. “This is such an unlikely and serious flaw,” computer scientist Toby Walsh told the BBC, because Java is designed to handle memory safely. Oracle and OpenJDK have released fixes.
58
seo ruined search
Accept the reality that Google AI answers are for many searches an improvement for the user. Would not be maybe in the web of 20 years ago. Now it was a shit party of useless sites, advertising of all the kind, and other garbage.
35
i'm surprised nobody gave them a heads up that the model changed right before embargo lifted
Replying to @MiaAI_lab
Yes, that's the whole point of getting feedback
36
where does neon even sit between hydrogen, carbon, and argon?
Gemini 4 Neon is coming out today then It's listed in the Google cloud billing catalogue, what on earth is going on
2
3
358
Zack Swafford retweeted
Claude Fable 5.2 is in Dart's model picker.
1
41
bullish on cybersecurity
5 MAJOR DISRUPTIONS IN JUST 6 DAYS 🇺🇸 Nov 17: NSA classified network DOWN 🏦 Nov 16: JPMorgan banking failure ✈️ Nov 16: United Airlines outage 🛑 Nov 13: FAA nationwide ground stop 🌐 Nov 12–13: Microsoft 365 outage No coordinated cyberattack confirmed.
1
2
34
raising $5b six months after the last round means they must be seeing something really special internally.
Thrilled to announce $5B in funding for @IsomorphicLabs to accelerate our mission to Solve All Disease. We continue to make rapid progress reimagining drug design from first principles with AI to radically cut the time & cost to find new medicines. Gradually, then all at once...
1
2
49
doubling the previous best model in one drop
Talking about acceleration: In 13 months, the best AI agent went from completing 2.5% of real freelance projects on the Remote Labor Index to 41.3% with Claude Fable 5.2. These were paid projects that took human professionals 29 hours on average, with briefs provided and client calls excluded. So yeah, white collar work will get a wake-up in 2027.
1
1
26
cognition has been quietly shipping bangers lately
Introducing: Devin Apps Private, secure, sharable web applications running directly on Devin’s cloud computer. Devin Apps are designed to power your most ambitious internal tools, preview deployments, and extensions to Devin itself.
1
1
38
can't wait honestly
Effort levels (low/medium/high/max/etc) will disappear, just like planning mode. Most devs pick the wrong one most of the time anyway (esp. those who yolo on high). The agent should know what the task needs, and humans should seldom have to tweak the level.
1
1
38
Is claude down for everyone else?
204
wow, it actually happened
*NVIDIA'S JENSEN HUANG OUT AS CEO
1
37