Daniel R. Dafoe (@danieldafoe) header
Daniel R. Dafoe (@danieldafoe) profile picture

Systems, photography, fitness. ex: @apple @rbc @gbpolytech

Earth
Joined April 2010
Photo by Daniel R. Dafoe (@danieldafoe)Photo by Daniel R. Dafoe (@danieldafoe)Photo by Daniel R. Dafoe (@danieldafoe)Photo by Daniel R. Dafoe (@danieldafoe)Photo by Daniel R. Dafoe (@danieldafoe)Photo by Daniel R. Dafoe (@danieldafoe)Photo by Daniel R. Dafoe (@danieldafoe)Photo by Daniel R. Dafoe (@danieldafoe)Photo by Daniel R. Dafoe (@danieldafoe)Photo by Daniel R. Dafoe (@danieldafoe)
Daniel R. Dafoe retweeted
If you're interested in WebMCP, you should be following Idan Levin. He and his team have been doing extremely exciting work in this space- benchmarking, cataloguing, you name it! Strong suggestion to follow him.
Replying to @rauchg
You might find this interesting - We created the first WebMCP benchmark, comparing it with computer use and DOM-based methods across speed, cost, task success rate, and token usage. The results are incredible. webmcp.com/benchmark You’re getting 3-5x speed, and an order of magnitude cost improvement (up to 23x lower costs) while being 10-15% better at solving tasks compared to other methods
20
33
2
840
194,850
Daniel R. Dafoe retweeted
Replying to @isaac_ts_way @thdxr
A already has a metric we call wall clock time and B is decode tokens per second at the engine level. Including prefill time in decode tok/s isn’t helpful because it conflates two workloads. Instead you should form the whole picture with many metrics: TTFT, decode TPS, prefill time, and wall clock time which includes e2e latency including HTTP overhead
1
1
1
3
186
Daniel R. Dafoe retweeted
New OG images on ogpedia.xyz: 01 • aquaduck.ai 02 • telemetry.dev 03 • letude.group 04 • modal.com
3
20
1,191
Daniel R. Dafoe retweeted
Lot of good points but also fair bit of confusion on what shared compute gets you so let’s clarify 1. Shared compute does NOT mean model gets smarter over time. Weights are frozen at inference time. You’d need something like shared memory layer or plan to post train on data 🧵
I think the most interesting thing about Jack Dorsey's "Slack killer" is the idea around shared compute. I haven't seen people talk about it so here are my thoughts FWIW: Open models got good, close enough to the paid frontier stuff to run for real. But the strongest ones need expensive hardware most people probably won't buy alone, and it's kinda a pain to set up if you aren't technical. Shared compute solves exactly that. In Buzz, one person runs the machine, loads up an open model like Google Gemma, and everyone in the community plugs into that same model. Basically, a whole group has real AI they own and control together, running on their own hardware, learning from their own data. Once you see it, a bunch of things click into place. 1. A community can now run a top open model together, on a machine they own, instead of renting from a lab. 2. It learns from the group's private data and gets sharper over time, and all of that stays inside the community. 3. A narrow, private model can quietly get better than ChatGPT for the one world your group lives in. 4. It's impossible to copy, because the edge is the private data on your machine, not the model itself. 5. The moat stops being how smart your AI is and becomes whose data it learned from. 6. Compute becomes something you share like a building shares a gym. 10 people split one machine instead of 10 people each renting forever. 7. Idle compute becomes income!!! Your machine sits dead half the day, so it earns money renting that time to someone who needs it. 8. Communities become the unit of intelligence instead of companies. The group with the smartest shared brain wins, and being a member means owning a piece of it. 9. A shared brain becomes an asset you build equity in. You put in money and data, it appreciates, and your slice is worth something the day you leave. 10. The whole thing runs on open protocols, so the group keeps full control and nobody outside can throttle it or shut it off. You know me, obviously, my head went to what startup ideas come to mind here. Adding them to @ideabrowser soon. Well… 1. The vertical brain. Pick one profession, tax lawyers or real estate agents or indie game devs, and build the shared machine trained on everything that group knows between them. A year in it's the smartest AI in that field, impossible to copy, and you own the club it lives in. 2. The rental marketplace for collective brains. Once these private models exist, outsiders will pay to use them. You build the layer where a group lists its brain, an outsider pays per task, and the money flows back to the members while you take a cut. A marketplace for expertise, not compute. 3. The idle-compute exchange. Every shared machine sits unused half the day. You build the market that rents that dead time to whoever needs the power right then, so owners earn money off a machine that was just sitting there. Idk where Buzz goes, but it's cool to see Jack putting it out. Right now the way it works in AI is you rent your intelligence from a few giant labs that own the machine, set the price, and hold the off switch. Shared compute flips that, because a community can run the model together, feed it their own private data, and keep full control of the whole thing. It's one of those things that might look tiny today, but Jack does has a habit of being early.
2
5
1
11
3,148
Daniel R. Dafoe retweeted
Day 2 of #a11ytoconf is here! Ready for more accessibility insights, inspiration, and innovation? Let’s do this! 💪
1
2
77
Daniel R. Dafoe retweeted
as a senior engineer you can be kind, you can be right, or you can be:
5
10
37
8,587
Daniel R. Dafoe retweeted
Maybe I’m old or maybe I’m young but I don’t understand what the problem with the question is here
This is a very on-brand Hacker News comment
3
1
2
378
Daniel R. Dafoe retweeted
🎉 We're thrilled to have @techgirl1908 & @kirahsapong leading a workshop on tbDEX at this year's #fintechdevcon! If you're building #payments, this hands-on session is for you. 💪 More about Angie. 👉 loom.ly/aQYM3xM More about Kirah. 👉 loom.ly/Bjyhuws
5
11
2,711
Daniel R. Dafoe retweeted
See you at @fintechdevcon!
I was so happy to learn that there's actually a conference specifically for fintech developers, and even more excited to teach a workshop! If you're in fintech, you should definitely come to @fintechdevcon fintechdevcon.io/
4
9
3,955
Daniel R. Dafoe retweeted
ARKit is teaching me more about biology than biology ever did
1
1
72
Daniel R. Dafoe retweeted
The W3C seeks an Accessibility Specialist, US remote #a11y w3.org/careers/2023-accessib…
1
15
16
2,046
Daniel R. Dafoe retweeted
thanks to this game i just lived long enough to see myself become the villain shadysam.com/
1
1
49
Daniel R. Dafoe retweeted
“the design is so lazy” to YOU but my 19yr old disabled sister who can’t put shoes on w/o assistance would LOVE these bad boys
Nike has unveiled their first hands-free shoe: the Nike GO FlyEase.
1,439
100,990
2,007
793,718
Daniel R. Dafoe retweeted
dressing cannot feel pain
This Post is from an account that no longer exists. Learn more
20
234
7
2,144
Daniel R. Dafoe retweeted
not a jug of dirt
14
58
9
657
Daniel R. Dafoe retweeted
bottoms also plain
31
247
89
1,646
"eleven plus two" is an anagram of "twelve plus one".
4
9
31
Daniel R. Dafoe retweeted
The joke's on you, companies! My resolution this year is to read every last word of every Terms and Conditions I encounter. #NewYearsResolutions #ToWhatEnd
1
1
1
Daniel R. Dafoe retweeted
I forced a bot to watch over 1,000 hours of stand-up comedy and then asked it to write stand-up comedy of its own. Here is the first page.
1,480
17,279
3,364
62,039