Netcup - connect to tailscale - deploy updown, coffeechat, factory, cvcl OpenClaw - setup to WhatsApp with Rick and Morty The EHT image should be black and white and should look drawn though. The main part of the picture should be the spacetime diagram geometry (the right bottom, and it should contain more stuff). The labels should be understandable for cambirdge 1st year undegraduates in physics (explain what is Schwarzchild radius and so on). The cones on the spacetime diagram are not right, also do no include I and II and note down event horizon in that diagram. It should also contain x-axis as position. The title should be Black holes singularity. I want you to make another deisgn called "Kelvin-Hemholtz instability" or "Turbulence instability". It should show video snapshots of how the instability develops in 3-4 pictures. There should be labels explaining what is going on with the pressure and what is the scenario, the last snapshot should be saying "It is everywhere!" and show how it is for example in the big eye of Jupiter, clous and so on. The pictures should look like it is scketched and again, the labels should be understandable to first year physics undegrad. ------------------------------------------------------------------------------------------------ Stash - QED higher order diagrams - Pulsars - Fermi Surfaces? - black front with logo, back with concept and sponsors Gift for the speakers About the disc: - basic folder - logo, constitution, list of people, termcard - speakers - proposed, emailed, template, record - financial record - sponsors - stash - archive - photos ------------------------------------------------------------------------------------------------ judge it by the view Tripos is illusion, you should not lean on it. The most exciting part are the things you make yourself. - reach out to cool people you want to work with - CUPS - the goal is to work for yourself with cool people, do not beg other people for work. - AI - use it - be humble Top 10 men experience - watching your girlfriend getting ready in front of the mirror - Christmas - watching fight club for the first time - going to play football in your home village in the summer - leaving lib in 1am - shit in the empty house - swimming in the sunset ------------------------------------------------------------------------------------------------ QM - randomised benchmarking - strat is return instructions (could be circuit) and then process results for BOTH verification and calibration - Is there a way I can fake run it? - If I am running only one qubit, how do I know the E or O. - Should I leave QPU config? - Should I think about check time or cal time? - artefacts - CheckPointBase - Evaluate metricss - Problem with Bad_Data in check_data is how we gonna tell that calibration fails? Chain out_of_spec, calibration fails -check data return all three of them if it can -calibration return file or None. if none than it Runner returns False in runner. If False than it changes CheckDataResult into Bad_Data i.e. passes down -Checkpoint same as CalibrationBase but ca but calibration returns None by default -constantly returning Out of spec does not work what if bad data -runner changes? - calibrate in runner returns true if successful, False otherwise - the returned metric should be in artefacts - any reason why calibration node not artefacts? - diagnose and maintain are almost the same - only maintain calls check_state before - for now still name identification for node -I think the node should be selfcontained and both Verification only and calibration should be the same class. There should be flag if checkpoint or calibration. - Maybe the naming should change, not IN_SPEC, OUT_OF_SPEC and BAD_DATA but based on what it should do i.e. RETURN, CALIBRATE, MOVE - things I should do: update verifications so it gives BAD_DATA if appropriate, last_artefacts, saving T1 etc in config file?, you should clean up the code and correct the tests - DFS should check if visited not only loop. I think check_state should not include recursion. The recusrion is only in maintain, not check_state. This improves from O(n^2) to O(n). - how to handle state if we actually cannot calibrate? - need to correct docs and tests Discuss today with Julio - Removing recursion in check_state - addint visited - tests with Claude - more tests with calibration - Runner should be running the Loop for LoopNodes # Most of the logic about being checkpoint should be in the Node and runner. # It has verification functions only and it can only return IN_SPEC or BAD_DATA # The relevant metric should be returned in artefacts # class CheckPointBase(CalibrationBase): - documentation - VerificationState vs CheckDataResults - clean up with the commits - I am not sure if you want to keep tracking visitors in diagnose - _log_action - override repetitions - weaker "correlation" edges is there any way we can have two qubits? - need to isolate its hidden link from hidden link in the dependencies. The current logs are horrible, we need to figure that out. - what return when calibration not implemented? - docs how to change repetitions - should the dag files be in local files or apollo. - we should be able to estimate expected calibration time - monte carlo - what is the point of the snapshot when it does not store its informations. what is the point of json - deterministic dependencies are ok - how to implement hidden dependencies? => first step -log(P(A and B)/P(A)P(B)) ig. what module what class? speed of the replay iterate qubits and qubit pairs iterate what confidence do we want to set up, to have as fast dag as possible? Very important, you can speed up your process by a lot. Same for OUT_OF_SPEC vs BAD_DATA have last_state in dag, so it can be loaded easily. On diagnose, I should check in order of increasing confidence (the least confident first). https://www.spareroom.co.uk/flatshare/flatshare_detail.pl?flatshare_id=16584026&search_id=1475682760&city_id=&flatshare_type=offered&search_results=%2Fflatshare%2F%3Foffset%3D30%26search_id%3D1475682760%26sort_by%3Dby_day%26mode%3Dlist& https://www.rightmove.co.uk/properties/91198662#/?channel=RES_LET I should simulate two things - with probability and without probability - compare time elapsed. For longer experiments, you should have longer initial confidence than for shorter. What graphs do I want: - time of run of calibration vs confidence required - what fraction of time we spend miscalibrated and run experiment at the same time? - how long on average does run take - how often do we need to run - how often we are in wrong - what percentage of my runtime is taken up by my calibration - what percentage of my efficient experiment runtime is miscalibrated - how to get the most runtime? calibrate in order of confidence. human check png show up code somewhere nicer errors Maybe we can gamle much more - get the chain of the weakest link - do some binary search on it. Maybe count how many subdags exists by removing some nodes. Human Mode. Add continuous simulation mode. machine learning transformer. How does gitlab work? If dag is going to scale up, is it going to need to train again and again. Ideas from you, code from claude The probabilities should be displayed over the whole time of the simulations for all the nodes. We always know the probability of being micalibrated as described in the pdf (read the section again). You should compare the probability I believe based on my mesurements (estimated probability distribution) and the real internal probability (real probability distribution). You should display the simulated time and how it changes over the simulation somewhere (clock). Also some relevant scales/lifetimes for the nodes - barthel 24h, frequency 30mins, ramsey 20mins, amplitude 40minutes, Cz 240mins, theta 240mins. At that time the probability of being miscalibrated is 0.5. Next, I want you to have some general statistics about simulations somewhere (above the node section, bellow the dag title) - histogram distribution of the times of the simulation including mean and std. I want the simulations to run in batches, so I can change the settings. I also want a mode of continuous simulation - firstly it is going to run the algorithm as usual, but then when the top node is calibrated, the clock just keeps running. When the probability of the called node decreases under 90% (or our own set treshold), it should run the algorithm again continuing its clock. It keeps going like this until the simulated time reaches some time (for example 24 hours). lifetimes are basically tresholds for probability. https://docs.google.com/presentation/d/1IxMrn7UasTW0gTH_hWr3-4SWMQ0-Tk-nbzalRb3CWnA/edit?usp=sharing Trading machine for plane tickets LinkedIn from Quantum Motions Finance app Habits Zurich to see my girfriend from Aug 12-16, cute dates, traveling around the town and data please. Up and Down hedge fund Books Open Library APIs AI optical table Book flashcards, paper flashcards From Imamoglu: Theory difficult but interesting, experiments simple idea, diffucult to do, but interesting Revolve anyons around each other and do interferometry (channels) Quatum Hall Effect Ultracold has to be overstaffed more complex Limitation is manufacturing Microscope incorporated Transition preserve topological invariance Excitons as way to sense stuff. From Esslinger: Zoran is good but not easy Higher order beats and phase space seperation. Lithium, cavity, lattice, lattice and Cavity 60 Hierarchy in Zuruch 8 ppl is too much Time variation changes population and it gives you Heisenerg limit. Momentum squeezing From Wallraff: Achieved 3 surface codes, google wants depth 5 soon. Still above treshold. No hero device. Wet vs Dry attenuator and aplifier (schemes) 30m long tube with cooling in the middle. Cheating by shifting equipment. Certifiably random. Loop-free Many officials Proud intense but well organised 250 repos Chrystanos Travel Planner - rid of chat, expand window to the right, chat window should be only up. Alex narrvall- pravda, život, instagram, vape Music video with formal and techno Media maths models about information traveling social media, wherenext.com I want to travel to Zurich to see my grifriend. I was thinking dates of 20-26 Aug but this can be vaguiely flexible. I am traveling from London to Zurich and then from Zurich to Prague. Acommodation is provided. While there, I wanna plan out some dates and visit ETH Zurich. Also some swimming spots if cool . 2% transanction fee from coinbase to wallet (including network and exchange fees). Finance: http://100.97.215.52:8000 Travel - better plane animations, have you ever ...? finance, travel, emails, cups, health, apps, thoughts, VM, IoT, books, education Hackathons - Machine learning, AI https://id-preview--84a7cc46-2c41-4e74-b447-6c043e1f717e.lovable.app I see signs of you on every corner. NAMES jdej.dev, art, traphouse.dev coffee chat - email app factory - builder up-and-down, updownfund - finance cups - cups wherenext - travel meditations - thoughts health - jordan shopping list - robe, bowl, pan, other cooking stuff, boxes for food, and so on. TO DO - cooking, quant, BEC, house, cups, home email - eugene, iter guy, caltech girl, tong and so on (+hannah fry) jane street - merch, message them 1650 sponsors - make a chat, put tiers, potentials and current on the drive, rough term card where they can put our things. hackatlon, ask mathmos about jane street, find the old spreadhsheet. https://claude.ai/code/artifact/d931feb7-1d41-4439-ae07-ca3ee81d2a56 tourist in his own life TO DO today: email St Johns and Caius jdej.dev cvcl Green Book Netcup
export
saved
+-- TO DO --+
saved
(empty)
>