==== Ivan Krstic, the OLPC project ====

- Learning: peer, self-driven until age 5-6, then based on authority (works with great teacher)
- Do laptops now a let schooling fix itself over time
- Computing power comparable to year-2000 workstation
- Dual-mode display: medium-res color with backlight, very hi-res BW without (sunlight readable)
- Mesh networking active during suspend
- Touchpad is also tablet
- All open source
- Sugar GUI emphasizes collaboration and simplicity
- Bitfrost security platform: virtualization, no user involvement
- Python everywhere: from educational philosophy (also, defeating complexity with beauty)
- What should the learning to learn be about?
- MetaSci (from MetaSploit): a framework to make visualizations/Hypercard-style about parts of datasets, scientific results
- This seems compelling enough for people to roll demo sets

=== Travis Oliphant, Python buffer interface ===

- Getting numpy into Python
	- Get the numpy "structure" into 3.0 via the buffer interface
	- Numeric, numarray, numpy all use the array interface to share data

=== Chris Mueller, Python on IBM's Cell/Broadband Engine ===

- Cell BE: heterogeneous multicore processor -> 2 general-purpose PowerPC, 8 special-purpose Synergistic Processor Element
	- Three instruction sets (PPC, Altivec, SPU)
	- Programmer managed local store on SPUs
	- In order execution on all cores
- SPE details
	- 256k local store
	- 128 x 128bit registers
	- multiple communication channels
	- better double precision coming...
- programming models
	- Manager/worker
	- Pipelined execution
	- SPE threads (processor in memory)
- Cell/Python programming model
	- Python for low-performance tasks on PPE
	- Native kernels for high-performance tasks on SPEs
- CorePy
	- Create and execute PPC, VMX, and SPU programs from Python
	- Execution Management and Code Synthesis components

=== Enrico Vesperini, Beowulf Analysis Symbolic Interface (BASIN) ===

- NSF-funded, 5 years, Steve McMillan
- Stellar dynamics, n-body simulations

- Beowulf clusters are attractive, but parallel data analysis is difficult
- BASIN: suite of tools for parallel data analysis and visualization
- C++ library, but use interactively from Python

- start GUI, connect to remote servers, read file

- lots, see website