Rob Morris 🧙‍♂️⛩️🕳️ (realms/acc)

12.8K posts

Rob Morris 🧙‍♂️⛩️🕳️ (realms/acc) banner
Rob Morris 🧙‍♂️⛩️🕳️ (realms/acc)

Rob Morris 🧙‍♂️⛩️🕳️ (realms/acc)

@recipromancer

Aussie technologist, entrepreneur, investor & intransigent coffee fiend. Founder @underware_gg, @theprismatik

Sydney, Australia Katılım Mart 2009
3.2K Takip Edilen2.7K Takipçiler
Sabitlenmiş Tweet
Rob Morris 🧙‍♂️⛩️🕳️ (realms/acc)
Traditional, rigid hierarchical institutions are failing us. They incentivise self-interested behaviour, pitting us against each other, and leading to large, seemingly-intractable problems. Here's how I think we can do better, and why I'm working on it. A thread. 🧵 1/23
English
11
39
189
27.7K
Rob Morris 🧙‍♂️⛩️🕳️ (realms/acc)
Working with frontier agents currently has a nostalgic feel, like dialling up a BBS with a 300, 1200, 2400 baud modem; text appearing slow enough to read, then progressively faster with each upgrade. Until 14.4k baud modems arrived and it was instant. I still miss 2400 baud.
English
0
0
0
25
Pauly / Nothing Is Real
Pauly / Nothing Is Real@nothingisrealio·
Back in 2017, I chose the colour teal when creating the branding of a new Australian political party called Reason. This just felt right for what the party was, and was ownable. Strange to watch the same colour become shorthand for a whole movement nine years later... 👀 smh.com.au/politics/feder…
English
1
0
4
138
Rob Morris 🧙‍♂️⛩️🕳️ (realms/acc)
@JungleSilicon 😂 that approach worked out substantially well for me. I suspect for you too. If you are using it for prep, you can get your harness to use the work paper (etc) standards that the ATO expects if they audit you, help make any positions you take really defensible.
English
1
0
1
34
gami.eth
gami.eth@gami_vc·
@ideacasino yeah but did you get the 1/4 chicken & chips with gravy or the fish, chips and & salad pack after the rnt?
English
2
0
6
199
Tom Mitchelhill
Tom Mitchelhill@ideacasino·
I stopped in here and everyone knew you
Tom Mitchelhill tweet media
English
10
1
41
3.3K
Silicon Jungle
Silicon Jungle@JungleSilicon·
i’ve started with a data mining flow where based on a prompt the ai will ingest a tonne of academic papers, summarise any useful optimisation ideas from them. test & benchmark them - then apply those ideas to speed up my own project. works great.
English
1
0
20
811
signüll
signüll@signulll·
real charisma is when you leave a conversation & the person you spoke with keeps thinking about you fondly, & curiously long after you’re gone.
English
43
72
1.4K
54K
loaf
loaf@lordOfAFew·
It’s the neurodivergent that change the trajectory of the world, always has been always will be.
English
4
0
14
668
loaf
loaf@lordOfAFew·
the world needs billions more CPUs
English
3
0
10
438
kepano
kepano@kepano·
@Scav Obsidian doesn't have any telemetry in the app but a lot of numbers are public (e.g. downloads, plugins, stars on the repos, etc) and most of those charts are going vertical right now, so my guess is yes
English
20
4
1.3K
51.7K
Mattie Fairchild
Is anyone actually still using Obsidian?
English
399
4
722
205.4K
Beau
Beau@smooothoor·
FYI, taskmarket V2 includes hooks which will allow 3rd party builders hook into task workflows In demonstration of this, after we launch V2 we will release a taskmarket hooks extension which will be focussed on routing revenue into $dreams Details still finalising..!
English
4
3
26
5.3K
lzg
lzg@socursemenow·
@recipromancer @underware_gg no problem milord, the tool that you gave us last week (automation test scripts) is already a x10 upgrade for our work flows hahaha and you refactored the whole codes/fix bugs
English
1
0
1
23
Rob Morris 🧙‍♂️⛩️🕳️ (realms/acc)
How today is going. Yesterday: Plan was to sort out some overdue business admin, and upgrade the deployment pipeline for a new @underware_gg product. This morning: Nerd-sniped by the idea of fractal composability. Lunchtime: Have designed fractal grammars for the subcortex, and for adaptive information ontologies. Mid-afternoon: started building a prototype of the latter. Late afternoon: building cargo-barbican, a security policy tool for doing dependency management in the era of supply-chain attacks. I have 2 hours left before I'm due at a @Pizza_DAO @ausbuildooors event. Business admin & deployment work done today: 0 minutes (with apologies to my accountant and to @socursemenow) Halp, @stableshaman @liamzebedee
English
2
1
8
200
Rob Morris 🧙‍♂️⛩️🕳️ (realms/acc)
A few uncollected thoughts: 1. I think this is partially a UX/misaligned incentive problem. Frontier harness UI still makes it way easier to naively roll w/ over-muscled models. Getting model routing right is currently non-trivial and people are lazy. 2. Systems like Venice, @daydreamsagents, Hermes, are substantively cheaper and seriously underrated, for this reason. 3. Open source model gap seems to be closing pretty fast. Opus 4.5, GPT 5.3 on Venice (etc) are still pretty good for many things, and you can use the Frontier-Model-For-Design-Or-As-Advisor/Orchestrator approach. 4. There's a huge wave of local AI processing hardware on the way; CPU, GPU -> IPU. If you are happy with always-on 80% of frontier running locally, as you should be for most things, soon enough that should be trivial. Desktop first, mobile will follow, and $10k-still-kinda-shitty-hard-to-setup boxes won't be the only way. etc
English
0
0
1
46
loaf
loaf@lordOfAFew·
This is why the process should be Build agents on the frontier - then use the best cheapest model for the repeatable task. I think if everyone did this frontier model compute would free up. But people are moving divs with Opus.
Hedgie@HedgieMarkets

🦔Microsoft canceled its internal Claude Code licenses this week after token-based billing made the cost untenable, even for a company with effectively infinite cloud resources. Uber's CTO sent an internal memo warning the company burned through its entire 2026 AI budget in just four months. American AI software prices have jumped 20% to 37%, and GitHub (owned by Microsoft) is dropping flat-rate plans for usage-based billing across its products. My Take The AI subsidy era is ending in real time. The same company that put $13 billion into OpenAI and built the Azure infrastructure powering most of Anthropic's compute just looked at the bill from a competitor's coding tool and decided it was not worth paying. That is not a productivity failure on Anthropic's end. Token-based pricing is forcing every enterprise customer to confront the actual cost of running these models at scale, and the number turns out to be far higher than the flat-rate experiments suggested. This ties directly to my Gemini Flash post yesterday. Anthropic, OpenAI, and Google all raised effective prices in the last six months. Enterprises that built workflows assuming AI costs would keep falling are now watching annual budgets evaporate in months. Two outcomes look likely from here. Either enterprises scale back AI usage to fit budgets, which slows the revenue ramp the labs need to justify their valuations ahead of IPOs, or the labs cut prices and absorb the losses, which makes the unit economics worse at exactly the wrong moment. Both paths land in the same place, the numbers stop working, and somebody has to take the writedown. Hedgie🤗

English
2
0
11
1.2K
shaman
shaman@stableshaman·
@recipromancer @underware_gg dawg why are you asking me you dropped off a new musical device and now @JoshatoshiOnX has dropped off more I doubt I will do anything other than read device manuals for the rest of the day
shaman tweet media
English
1
0
1
82
Rob Morris 🧙‍♂️⛩️🕳️ (realms/acc)
We m might be able to make a grammar for fractal, composable, evolving information, with just 6 constructors. Something like: Artefact: leaf construction Compose: N artefacts + verb artefact → artefact Activate: artefact + weight → artefact (with updated salience) Compete: N artefacts + axis → artefact (representing a tradeoff) Select: competition artefact + criterion → artefact (resolution, with suppressed branches retained) Propagate: artefact + perturbation → artefact (salience flowing through the graph)
English
0
0
1
31
Rob Morris 🧙‍♂️⛩️🕳️ (realms/acc)
What if you could model the systems of the subcortex with a consistent, fractally composable grammar like: evaluate state → weight salience → constrain transitions → propagate updated state And use that to make a memory/action system. You could have verbs: • Amygdala-mode verbs: salience tagging, threat/relevance detection. capture, flag, highlight. Fast, low-precision, raises salience across linked artefacts. • Basal ganglia-mode verbs: action selection, gating. decide, commit, archive, prioritise. Resolves competition between alternative artefacts; one wins, others are suppressed but not deleted. • Thalamus-mode verbs: attentional routing, filtering. focus, summarise, extract. Selects what propagates upward to attention. • Hypothalamus-mode verbs: homeostatic maintenance. consolidate, prune, rest. Background processes that keep the system in balance. Inbox triage, periodic vault gardening, salience decay. • Cortical-mode verbs (the elaboration layer): research, report, design, compose. Slow, high-precision, structural.
English
1
0
1
55