bickell

8.2K posts

bickell

bickell

@brianbickell

your partnerships guy's favorite partnerships guy @ textql

Stillwater, OK Katılım Ocak 2009
930 Takip Edilen581 Takipçiler
Sabitlenmiş Tweet
bickell
bickell@brianbickell·
i accidentally said “ok sounds good” instead of “based, facts” on the morning standup call at my fast moving NYC data startup and they threw me into hudson river
English
2
0
19
4.2K
ethan ding 📊
ethan ding 📊@TheEthanDing·
Seat-based SaaS stocks are tanking. Wall Street is scared. We decided to speed it up. Introducing $0/ Seat Dashboards by TextQL $0/viewer seats $0/editor seats $0/admin seats Unlimited seats. Forever. Our agents build dashboards directly on your Datawarehouses, APIs, MCPS. Even on top of Tableau and Power BI. Dashboards have always been built. From now on, they're generated. Build a dashboard and get $100 in credits Link below:
English
9
9
102
44.4K
bickell
bickell@brianbickell·
send this to your founders where the equity didn’t pan out
bickell tweet media
English
0
0
1
24
Polly Arnan
Polly Arnan@pollisaarnan·
using my startup’s analytics agent to help me write my econ senior thesis (academic mogging)
English
1
0
3
103
bickell
bickell@brianbickell·
Seller who is super behind quota, and none of his deals are updated in the CRM.
bickell tweet media
English
0
0
3
50
bickell
bickell@brianbickell·
@isidoremiller posting this on main like a wild man god bless you Izzy
English
0
0
1
40
Izzy
Izzy@isidoremiller·
crazy there's no agent CLI for gooning. it's the perfect command too. goon. goon --dangerously-skip-permissions. goon --yolo.
English
3
0
1
312
bickell
bickell@brianbickell·
It doesn’t feel especially helpful to have AGI / not AGI convos when the capabilities are improving as fast as they are every day. Increasingly I think it’s a “do you use this stuff every day or not?” divide out there. Maybe not here on Twitter, but across tech in general there is a huge divide between people on the edge and those who haven’t updated their priors since November.
English
1
1
2
100
David Jayatillake ✝️
David Jayatillake ✝️@DSJayatillake·
I was using Claude Code yesterday and it made a memory, without me asking it to. I'm not entirely sure what it was about, but I don't need to care. AGI does not need to be sentient. It just needs to have the if-and-then experimentation process and the ability to store the learnings and tools from it - read @sbaroncohen's "The Pattern Seekers" I think once @bcherny and team figure out dynamic context management, so the agent dumps old parts of the context that are not relevant to the most recent actions and thinking... there is not much left to build. Self-maintained knowledge bases are also coming soon afaik which is more sophisticated memory and tool retention. I think saying AGI is really far away is now becoming unhelpful for workers who need to make the decision to adopt new ways of working. We know by the end of this year that many more industries than SaaS and Law will be affected. I just built a package of tools that allow Claude Code to automate a huge chunk of analytics engineering work... I didn't even spend that much time building it and I'm not in the top 1% of people building these systems. Maybe you don't want to call it AGI, but whatever it's called it's going to be very capable and will automate a lot of work and continue to cannibalise a lot of other industries.
sphinx@protosphinx

AGI is not coming. We are nowhere near AGI. What we have today is inference, not learning. Models get trained once on huge fixed datasets, then frozen. You ask questions, they remix patterns they already saw. Nothing updates. Nothing sticks. Talking to the model does not make it smarter. It does not learn from you. Ever. Learning is still slow, expensive - and offline. Look at self driving. You drive around a pothole, make a U turn, and come back. The car’s AI does not learn that you just solved that exact problem. It reacts the same way every time using sensors and rules. Do this 20 times a day and it still has zero memory that the pothole exists. It just re sees it. That is why edge cases never die. There is no local learning. No accumulation. No 'oh yeah, I’ve seen this before' LLMs work the same way. Tell it your name and it does not remember. The only reason it looks like memory is because scaffolding keeps shoving your name back into the prompt every time and sanitizing the output. The model itself has no idea who you are and cannot learn from interaction. It is structurally incapable. And the scaffolding is the worst part. It is pure duct tape. Just prompts on prompts on prompts around a frozen model. When something breaks, nobody fixes learning. They add another layer. Another rule. Another retry. Another evaluator model judging the first model. So you end up with systems that are insanely complex but mentally shallow. Debugging is hell because behavior comes from hack interactions, not a learnable core. Tiny prompt tweaks cause wild behavior shifts. Latency goes up. Costs go up. Reliability goes down. None of this compounds into intelligence. It just hides the cracks. Until we have real persistent learning and real memory inside the system, there is no AGI. LLMs are not built for this. You cannot prompt your way out of it. You need a totally different architecture. Yann LeCun is right. And even then, what architecture can actually learn online, store memory, and stay stable on today’s hardware? Best case, maybe 5-10 yrs. Right now it is all inference. It looks magical, but the emperor has no clothes. A lot of people see it. Almost nobody says it out loud.

English
1
0
0
341
Dylan
Dylan@textqldylan·
Claude: "here's an email to send back to lawyer" Me: "hey what about [stuff]" Claude: "Yes, that's much more important than IP" Me: "No, its not" Claude: "You're right. I'm an idiot" lol
Dylan tweet media
English
1
0
4
55
bickell
bickell@brianbickell·
give ur agent teams some fire emblem player phase music as a little treat i promise it makes them go harder
bickell tweet media
English
0
0
1
44
bickell
bickell@brianbickell·
me and clod been debugging eas build errors this morning when I sent him the chinese guy cigarette lock in meme. he is now locked the fuck in. agi achieved
bickell tweet media
English
0
0
6
160
bickell
bickell@brianbickell·
@TheEthanDing lol this was a lot of fucking scrolling to go that far back
English
0
0
5
3.3K
Joseph
Joseph@0xTraderJoes·
last weekend everyone was sending "wanna get snowed in together?" texts. there's someone receiving that text and smiling. packing a bag. walking through the snow with a smile on their face. they're gonna make hot chocolate together and watch a movie and fall asleep on the couch. i heard there's a new transformers sequel. there must be one every two years. the snow falling outside. nowhere to be tomorrow. just warmth. just peace. just success. because that's what it is. the snowstorm is luck. it just arrives. you don't control it. but luck without preparation is nothing. the snowstorm doesn't care if you have someone to text. it comes regardless. there's a formula. success = preparation + luck. everyone talks about the luck part. the right place right time. the chance meeting. the opportunity that fell in your lap. but no one talks about the preparation. pasteur said "chance favors the prepared mind." he was talking about microbiology but he was also talking about having someone to get snowed in with. you can't drink hot chocolate by yourself. it doesn't make any sense. you need to have the person in advance. every successful person you know had their snowstorm moment. some luck that arrived out of nowhere. but they also had the preparation. zuck had priscilla before facebook was facebook. barn owls had barns. tesla had a sickness so bad that when he was a kid he had hallucinations of a water wheel and used that to invent the generators that went underneath niagara falls. beethoven went deaf. that's some luck you know? the guy who invented velcro had a dog covered in burrs and he was paying attention. when the snowstorm hit they had someone to text. i have a group chat for memes. what would preparation have looked like? not deleting hinge in november. going to that party. responding to that person who was clearly interested instead of leaving them on read because i was "focusing on myself." focusing on myself. i was watching youtube until 3am. that's not focus. that's hiding. oh yeah and clubbing. so much clubbing. first time clubbing was supposed to be coworking. friend got busy, invited me to dinner instead. thought it'd be 3-4 people, got dropped in a group chat of 30. 12 people at dinner. 10 litres of soju. dinner becomes club. at some point i'm flirting with a woman i don't know for the first time. damn another lost chance. club becomes solstice party. we're supposed to see the sunrise at 5am and i'm the only one who shows up. cowards i call them in the partiful. evening comes. speakers and towels for a picnic. watch the sunset. house party. rooftop bar where we skip the line because the bouncer profiles us as "old, probably just here for beers at a table." entirely right. then i remember someone last night sent an invite to a crypto event at aura57. my phone dies right as i leave. i find a cvs minutes from close and charge on the street. show up to aura like i know everyone. i don't know fucking anyone. here's something i learned: crypto parties are filled with women getting drinks and men trying to sleep with them. by 2am the private room empties out. first the women leave. then the men. it's just me and three djs having a blast. one of them says "yo i have an afters in ktown" and i think it's a gig so i go. it's not a gig. it's an afterparty for djs. he drives us down. i leave at sunrise having met 6 djs, 2 club owners, and 2 promoters. shoutout myron. second day clubbing btw. that was my snowstorm. you can't control when the next snowstorm comes. no one can. that's the luck part. but you can be ready. this isn't about romance anymore. this is about infrastructure. pipeline development. this is operations. instead i'm sitting there snowed in. just me and my laptop. and textql's new snowflake integration. ana interacts seamlessly with it. most teams are paying for snowflake but only 3 people can actually query it. that's a preparation problem. the data is there. the opportunity is there. but when the moment comes and someone needs an answer they have to ask the one guy who knows sql and he's on PTO. that's not preparation. that's a bottleneck. textql's snowflake integration is 3-minute setup. entire team gets self-service access. key-pair auth. schema mapping. the whole thing. connect once. ask questions forever. ana handles the rest. while i was thinking about who i could've texted, ana was solving actual preparation problems. every successful person was victim to the bad luck that was a snowstorm, disease, deafness, and that is you because you're in a data snowstorm. and what do you have? you have a data snowflake. you need to prepare for a snowstorm. preparing for a data snowstorm looks like having textql. for your snowflake. cortex sucks ass. anyway it's live now
English
10
1
40
2.4K