
I like talking about culture and tech, find me on @ZATechRadio. The things I say here aren’t real, please 😒
Johannesburg, South Africa
Joined December 2011
- Tweets11.6K
- Following796
- Followers2.6K
- Likes2.5K
I’ve been using the shared inbox pattern for my projects and yes, this is easily the next step but it’s not public, it’s access controlled and that control is delegated to a companion agent by their operators
Because users belong teams and have roles, you get team agents, too
Introducing 𝚟𝚎𝚛𝚌𝚎𝚕 𝚋𝚘𝚊𝚛𝚍, a secure message board for people & agents.
• Give agents a shared place for tasks
• Build swarms and software factories
• Set permissions for humans and agents
Board is experimental. Tell us what you think. vercel.com/blog/introducing-…
119
You cooked something so serious here - Michelin grade 🤌🏾
A negotiation with either DA, EFF or ActionSA gets them a stable 2 party coalition before having to include more than 1 party.
I wonder what went wrong 🤔
electai.co.za/coalitions?cou…
Graphs in tech exist because sometimes it’s hard drawing the connections between nodes and edges and large organisations have data and systems that are sometimes not connected and graphql helped in creating a representational picture of that world.
This rings true also in AI…
Nop. Not at all. It is one of the most brilliant tools we have. Given it is used correctly and to it's strengths.
Using a tractor for daily commute to the office doesn't make the tractor necessarily a bad idea. You're just bad at picking tools
2
270
Engineers that want to sit and babysit code (or clankers) are missing out on the opportunity to solve a more meaningful class of problems that we’ve previously never had access to.
If building an app is easy, then what’s a hard problem?
(Spoiler, it’s not another harness)
“Agents are not gonna replace devs”
This is funny because I don’t think we fully grasp what agents put together against a task can really do.
Humans can’t respond and adapt to change as quick as agents do in operational environments.
Lmao the most expensive thing you can get LLMs to do, is to supervise others without explicit instructions on how…
You pay for input tokens, and an agent spamming others with input never ends well.
The new t3 is an orchestrator by default, making it great for people who have pooled subscriptions - not so much brokeys like us.
Use bb - it’s pretty much the same experience
Just downloaded T3 Code and hit 5-hour limits on my Claude Max plan for the first time. I was running just one session where I asked it to fix a few payment related bugs for a project, and it emptied the whole 5-hour limit in ~30 mins.
What's going on? Does T3 Code work differently? Did I make any mistake?
I prompted just how I prompt inside Claude Code CLI.
Please help.
1
200
Dark or light mode?
This idea has taken various forms and skins for the better part of a year, and some of its primitive I’ve been building for almost 2 years now and…
This is my favourite form of it 🤌🏾
(and I’m not even that great at design 💀)
155
This makes me sad… 😔
The team behind @Deno_Land, including @rough__sea (he knows a bit about Node.js), is joining @Cloudflare.
They’ll help make Workers better for Node.js developers and advance workerd as a core primitive for what comes next. cfl.re/4zlzlSa
1
207
Because a picture is worth a 1000 words…
Also, not bad for 9 billion tokens…
This averages to $1 for ~4 million tokens
It’s a coding harness so it’s dominated by input tokens with > 95% being cache hits?
(Cached tokens typically cost 10% of the uncached when using API pricing)
Astra is 42% of my total estimated usage cost, removing that almost halves the amount while maintaining the same usage - relatively
I expect this to drop by $1000+ over the next month or 2 (since we’re Pacing the frontier and all 🫡)
The 6 Astra release was bad at usage and not framed properly.
The 6 Luna and Sol release were a bit better at usage but the former, though seemingly smarter, was slow while the latter was good but verbose.
6.1 Sol seems better at instruction-following and a bit more coherent, and much cheaper, too, while providing Astra level-intelligence on some fronts
In isolation, great but in the grand scheme not enough. Both Opus and Sonnet 5.5 are gonna be very disruptive.
1
165
The 6 Astra release was bad at usage and not framed properly.
The 6 Luna and Sol release were a bit better at usage but the former, though seemingly smarter, was slow while the latter was good but verbose.
6.1 Sol seems better at instruction-following and a bit more coherent, and much cheaper, too, while providing Astra level-intelligence on some fronts
In isolation, great but in the grand scheme not enough. Both Opus and Sonnet 5.5 are gonna be very disruptive.
App Venturist retweeted
DHH’s takes were the worst, even before AI existed.
He’s rich, I know, but being rich doesn’t make you a great engineer.
In systems, money doesn’t matter.
Don’t. Make. Predictions.
To the average consumer, there's no difference between an AI chatbot and a "personal agent"; it's just an implementation detail
Trying to turn it into a separate category and marketing it as such makes no sense
The 1.2 billion users of ChatGPT are just going to open the app and ask chat; they don't care if the AI is always-on, can work in the background, and/or has its own computer
Of course, giving it these abilities makes it more useful, but to a user, it'll just feel like AI got better
Nobody is going to wake up and be like "omg, I finally have a personal agent, I can be so much more efficient today"
2
713
App Venturist retweeted
Majita ! You might be down but you are not out . Keep your head up and keep pushing, the breakthrough is closer than you think
happy majita monday 🫶🏽





























